The digital asset market has fallen by 2.57 per cent to US$2.86T over the past 24 hours. Bitcoin, the largest token by market value, dropped 2.81 per cent to US$84,261.32 during the same period. The total crypto market cap declined 2.92 per cent, and Bitcoin’s move closely tracked that broader drop. The data shows a 96 per cent correlation with the S&P 500 and a 94 per cent correlation with Gold. Those numbers confirm that this move is not unique to crypto. Traditional markets and digital assets are responding to the same set of pressures.

Bitcoin’s drop triggered a leverage flush that cascaded into altcoins. Overbought conditions and a surge in derivatives open interest then amplified the pullback. The near-term outlook depends on whether Bitcoin holds above the US$2.76T market cap support, which sits near the 50 per cent Fibonacci level. A hold could open a rebound toward US$2.94T. A break below could extend losses toward US$2.65T.

The primary driver is macroeconomic. The Federal Reserve’s recent 25-basis-point rate hike and hawkish commentary fuelled concerns about further monetary tightening. At the same time, the 10-year US Treasury yield surged toward five per cent, its highest level since 2007. That move tightened financial conditions across the board. Strong US PMI data on September 23 reinforced expectations of persistent inflation and higher-for-longer rates.

As liquidity becomes less abundant, investors reduce exposure to risk-sensitive assets. Bitcoin behaved exactly like a risk asset in this environment. It sold off alongside traditional markets as participants priced in less liquidity. This macro backdrop matters because Bitcoin and other digital assets trade as long-duration risk assets.

When rates rise, the present value of future cash flows falls. Crypto does not have cash flows, but it still competes for capital. Higher yields make bonds more attractive. That shift reduces demand for speculative assets. The key items to watch are further statements from Fed officials and any movement in the 10-year yield. If that yield remains above 5 per cent, the pressure on risk assets could continue.

Also Read: Bitcoin’s US$87,000 spike: Real breakout or a US$900 million short squeeze?

A second force turned a measured decline into a violent flush. The initial macro-driven drop triggered a liquidation cascade. Data shows traders liquidated US$237 million in leveraged long positions in a single hour as Bitcoin broke below US$84,000. Over 24h, total Bitcoin long liquidations reached US$171 million.

Another measure shows US$158.95M in BTC long liquidations in 24h, a 243 per cent spike. Bitcoin dominance rose to 59.12 per cent as traders exited altcoin positions. This is a classic deleveraging event. Forced selling by overleveraged bulls accelerated the downward move, a typical sign of a crowded bullish trade unwinding. The scale of liquidations shows how crowded the long side had become.

A single hour produced US$237 million in long liquidations. The 24h total for Bitcoin longs reached US$171 million. The US$158.95M figure and 243 per cent spike confirm the same pattern. A stabilisation in funding rates and open interest would signal that the market has flushed out leverage. Until then, high liquidation volumes could point to further weakness.

The pain spread well beyond Bitcoin. Major altcoins underperformed the broader market. Avalanche fell 8.38 per cent, and Filecoin dropped 10.71 per cent. Both assets had enjoyed strong weekly rallies, with Avalanche up 36 per cent. That strength invited profit-taking.

The seven-day RSI for the total market hit an overbought 80.24. Traders rotated out of recently high-performing assets and into stablecoins or large caps. This rotation amplified the sell-off. Total open interest rose 11.13 per cent to US$493.14B even as prices fell. That combination indicates lingering leveraged positions that could fuel more volatility.

Avalanche and Filecoin had rallied hard. Avalanche gained 36 per cent in a week. That move left the market vulnerable. The 7-day RSI at 80.24 signalled overbought conditions. Profit-taking followed. Rotation into stablecoins or large caps is a defensive response. Sector rotation into stablecoins or large caps could continue if fear persists.

Also Read: Why did Bitcoin and Ethereum move in near-perfect lockstep after the Fed rate hike?

The near-term technical picture for Bitcoin now sits at a critical point. Bitcoin is testing the 23.6 per cent Fibonacci retracement level near US$84,432 after a rejection at the US$87,363 swing high. The structure remains corrective within a broader weekly uptrend of 10.56 per cent.

If Bitcoin holds above the US$84,000 support, it could retest US$87,000. A daily close below the US$82,000 to US$84,000 support band would shift focus toward the 38.2 per cent to 50 per cent Fibonacci retracement zone between US$79,600 and US$82,600. A deeper correction could reach the US$79,600-US$81,100 range. The US$84,432 level is the 23.6 per cent Fibonacci retracement. The rejection at the US$87,363 swing high set up the test.

The weekly uptrend remains positive at 10.56 per cent. A hold above US$84,000 keeps the US$87,000 retest in play. A close below US$82,000 to US$84,000 opens US$79,600 to US$82,600. The deeper zone is US$79,600 to US$81,100. The market will watch whether Bitcoin can absorb selling pressure and defend this zone.

The total crypto market cap faces a similar test. The key level is the 50 per cent Fibonacci retracement at US$2.76T. A hold above this support could lead to a rebound toward US$2.94T. A break below could extend losses toward US$2.65T. The pivot point sits at US$2.86T. Rising open interest alongside falling prices suggests that leveraged positions remain in the system.

The next 24h close relative to US$2.76T will matter. So will any shifts in spot ETF flow data. A rebound above the pivot at US$2.86T could target the recent high of US$2.94T. The 50 per cent Fibonacci at US$2.76T is the line. A rebound above the US$2.86T pivot could target US$2.94T. A break below US$2.76T could send the market to US$2.65T. Open interest at US$493.14B, up 11.13 per cent, shows leverage remains. ETF flow data is the next input.

My view is that this is a corrective pullback, not a reversal of Bitcoin’s strong weekly trend. The downturn has multiple drivers. Bitcoin liquidations started it. Altcoin profit-taking after a strong week worsened. The high correlation with traditional assets points to a macro-sensitive environment. Bitcoin and the broader crypto market remain connected to global interest rates and liquidity cycles.

For now, the evidence favours a liquidity-driven pullback, amplified by excessive leverage, rather than a change in the longer-term trend. Let’s see.

—

Editor’s note: e27 aims to foster thought leadership by publishing views from the community. You can also share your perspective by submitting an article, video, podcast, or infographic.

The views expressed in this article are those of the author and do not necessarily reflect the official policy or position of e27.

Join us on WhatsApp, Instagram, Facebook, X, and LinkedIn to stay connected.

The post Bitcoin’s corrective pullback or the start of a deeper drop toward US$79,600? appeared first on e27.

Singapore’s food waste problem is often framed as a consumer habit or a logistics issue. For Life Lab Resources, it is also a feedstock problem, and one that could help ease another pressure point in Southeast Asia: rising demand for more sustainable aquaculture feed.

The Singapore-based startup has raised US$1 million in fresh funding to expand its capacity to process treated food substrates and produce nutrient-rich feed for fish and other aquaculture species.

The round was led by Decarb123, which invested through DC123FW. Keng Eng Kee Holding also participated, alongside follow-on backing from AC S323, an angel syndicate led by Huang Shao-Ning.

Also Read: Winnow buys Lumitics as hotel kitchens turn to AI to cut food waste

The financing is modest by venture capital standards, but the signal is larger than the cheque size. It points to growing investor interest in circular-economy startups that can turn waste streams into industrial inputs, particularly in markets such as Singapore, where land, food security, and waste management are tightly linked.

For Life Lab Resources, the funding is meant to increase production capacity rather than finance a speculative bet. The company said the capital will allow it to process more treated food substrates and supply more aquaculture feed to farmers. It also plans to develop more circular food products, though the near-term focus appears to be on feed.

Why feed matters

Aquaculture is one of Southeast Asia’s most important food sectors. The region produces a large share of the world’s farmed fish and shrimp, and demand continues to grow as incomes rise and consumers look for affordable protein. But the industry has a feed problem.

Conventional aquaculture feed often depends on fishmeal and fish oil, ingredients made from wild-caught fish. These inputs are nutritious and widely used, but they are exposed to price volatility, supply constraints, and environmental concerns. Soy and other plant-based ingredients are common alternatives, but they come with their own land-use and nutrition trade-offs.

That is why startups and researchers across the region are experimenting with new inputs, from insect protein and microbial ingredients to agricultural by-products and upcycled food waste. The goal is not simply to make feed cheaper. It is to reduce dependence on stressed supply chains while maintaining the nutritional quality farmers need to raise healthy stock.

Life Lab Resources sits within this broader shift. By converting food waste into usable feed ingredients, the company is trying to solve two problems at once: diverting waste from disposal and creating a more circular input for aquaculture.

The model is especially relevant in Singapore. The city-state imports more than 90 per cent of its food and has made food resilience a policy priority. At the same time, food waste remains one of its major waste streams. Turning that waste into feed is not a silver bullet, but it fits neatly into Singapore’s push to extract more value from resources that would otherwise be discarded.

The circular-economy bet

Circular-economy startups often sound compelling on paper but can be difficult to scale. Waste streams are inconsistent. Processing costs can be high. Customers in traditional sectors such as agriculture and aquaculture are price-sensitive. A feed ingredient that works in a lab still has to perform reliably on farms, meet safety rules, and compete with established suppliers.

That makes capacity expansion an important milestone. If Life Lab Resources can process larger volumes of treated food substrates, it has a better chance of proving that its model can work beyond pilot scale. For aquaculture farmers, reliability matters as much as sustainability: feed has to be available, safe, nutritionally consistent, and sensibly priced.

Also Read: DELOS sparks ‘Blue Revolution’ in Indonesian aquaculture with Series A led by Monk’s Hill Ventures

The involvement of investors such as Decarb123 suggests an appetite for businesses at the intersection of climate, waste reduction, and food systems. Keng Eng Kee Holding’s participation is also notable because it ties the round to Singapore’s food and beverage sector, where waste is generated daily and circular models could eventually become part of operating practice.

Follow-on investment from AC S323 adds another layer of continuity. In early-stage climate and foodtech, repeat backers often matter because technical validation and commercial adoption can take longer than in pure software businesses.

Regulation will shape the pace

The biggest constraint may not be demand, but regulation. In Singapore, companies that manufacture feed for food-producing animals require a licence, and the Singapore Food Agency sets rules for some alternative feed inputs, particularly waste-derived materials.

These rules exist for good reason. Feed safety is directly linked to food safety. Inputs must be managed carefully to avoid contamination, disease risks, or harmful residues entering the food chain. For a company working with treated food substrates, compliance is not a side issue; it is central to whether the business can scale.

That regulatory burden can slow young companies down, but it can also become a barrier to entry once standards are met. In a sector where trust is critical, licensed and compliant operators may have an edge over informal or poorly controlled waste-to-feed models.

Singapore’s stricter environment could also become a proving ground. A startup that can meet the city-state’s safety requirements and demonstrate commercial viability may be better positioned to work with partners elsewhere in Southeast Asia, where aquaculture production is much larger but regulatory systems vary widely.

A crowded field, but not a settled one

Life Lab Resources is not alone in chasing the alternative feed opportunity. Across the region, companies such as Nutrition Technologies, Entobel, Protenga, and Inseact have built businesses around insect-based protein and other upcycled ingredients for animal and aquaculture feed. Globally, firms including Ÿnsect and Innovafeed have attracted significant capital to produce insect protein at industrial scale.

These are not like-for-like competitors. Some focus on black soldier fly larvae, others on specific agricultural by-products, while Life Lab Resources’s approach centres on treated food substrates. But they are all competing for a place in the same changing feed supply chain, one where farmers, feed mills, and regulators are testing whether alternative inputs can match conventional ingredients on nutrition, safety, cost, and scale.

For Life Lab Resources, that means the opportunity is real but execution will be unforgiving. The company must show that its feed performs consistently, that its supply of waste-derived substrate can be managed at volume, and that its economics work without leaning on sustainability claims alone.

Southeast Asia’s food systems are looking for practical fixes

The timing is favourable. Governments and companies across Southeast Asia are looking for ways to reduce food waste, improve food security, and lower the environmental impact of agriculture. Aquaculture is central to that discussion because it is both a major source of protein and a resource-intensive industry.

Yet the most successful solutions are likely to be practical rather than ideological. Farmers will adopt alternative feeds if they help maintain yields, protect animal health, and make economic sense. Food businesses will join circular waste models if collection, treatment, and compliance are manageable. Investors will stay interested if startups can move from promising pilots to repeatable production.

Also Read: Arus Oil is powering Malaysia’s circular economy by transforming used cooking oil into clean energy

Life Lab Resources’ US$1 million round does not answer all those questions. What it does show is that capital is still available for focused, infrastructure-heavy climate and foodtech companies when the problem is clear and the pathway to revenue is visible.

In a region where food demand is rising and waste remains stubbornly high, the idea of turning leftovers into feed is easy to understand. The hard part is building the system around it. That is the work Life Lab Resources now has more room to pursue.

The post Life Lab Resources grabs US$1M to turn food waste into aquaculture feed appeared first on e27.

Southeast Asia’s startup funding market is showing signs of life again, but the recovery is not reaching most founders.

Venture-backed companies in the region raised US$7.25 billion across 217 equity deals in the first half of 2026, according to the Southeast Asia Startup Funding Report H1 2026 by Kickstart Ventures and DealStreetAsia. That is the strongest half-year total since H1 2022 and nearly four times the US$1.86 billion recorded a year earlier.

On the surface, it looks like a sharp rebound after two difficult years. Look closer, however, and the picture is far more uneven. Deal volume fell 5 per cent year on year and is now nearly 62 per cent below the H1 2022 peak of 572 transactions. The number of equity deals is the lowest the report has recorded since 2018.

Also Read: Late-stage deals revive in Southeast Asia, but early-stage founders remain under pressure

The message is clear: capital is back, but it is not back for everyone.

A recovery led by a few large cheques

The rebound was heavily concentrated among a small group of companies. The five largest rounds accounted for 75.5 per cent of all equity funding, while the top 20 took 89 per cent. The remaining 196 transactions shared just US$800 million.

Much of this reflects a global shift in investor appetite. Capital is flowing into AI, compute infrastructure, semiconductors, robotics, defence technology and other areas seen as strategically important. KPMG recorded US$227.4 billion in global venture investment in Q2 2026, the second-highest quarterly total on record, with AI companies driving many of the largest financings.

Southeast Asia is participating in that cycle, but mainly through infrastructure and enabling technologies rather than a broad-based revival across consumer internet, fintech or software.

The clearest example is DayOne, the data-centre operator whose US$4.5 billion Series C round accounted for 62 per cent of all equity funding in the region during the period. The deal is tied to the global buildout of AI and cloud-computing capacity, where investors are backing the physical infrastructure required to train and run large AI models.

Without DayOne’s round, Southeast Asia’s equity funding would have stood at US$2.75 billion. That would still be 80.1 per cent higher than H1 2025, but far less dramatic than the headline figure suggests.

Minette Navarrete, President and Managing Partner at Kickstart Ventures, said the region is benefiting from investors seeking stability, diversification and supply-chain resilience. But she warned against mistaking capital concentration for market strength.

“Deal volume remains a better measure of market momentum, and on that measure, capital has returned but not broadly, and that tells us Southeast Asia’s recovery is still finding its footing,” she said.

Singapore pulls further ahead

Singapore once again dominated the region’s funding landscape. Companies headquartered in the city-state raised US$6.7 billion across 145 deals, representing 92 per cent of disclosed equity funding and 67 per cent of deal volume.

That lead partly reflects Singapore’s role as Southeast Asia’s financing hub. Many companies based there raise money to support regional or global expansion, not just domestic activity. Even so, the gap between Singapore and the rest of the region has widened sharply.

Indonesia, Southeast Asia’s largest digital economy by population and internet users, recorded only 17 deals and US$104 million in funding. No Indonesian company appeared among the region’s top 20 equity rounds. That points to an acute shortage of growth capital in a market that, only a few years ago, was producing some of the region’s biggest venture-backed names.

Malaysia showed the strongest breadth outside Singapore, with 27 deals worth US$203 million. Vietnam raised US$340 million across ten transactions, making it the second-largest market by value, although Vinpearl’s US$255 million round made up three-quarters of that total.

Also Read: When debt replaces equity: How SEA startups mask a funding winter

Thailand and the Philippines showed a similar dependence on single large deals. Amity Solutions accounted for about 77 per cent of Thailand’s US$130 million, while Salmon represented 75 per cent of the Philippines’ US$80 million.

Andi Haswidi, Head of Research at DealStreetAsia, said the US$7.25 billion headline should not distract from the weaker base underneath.

“Every market outside Singapore depended on one or two transactions to make its total. That is characteristic of what a market without depth looks like in any conditions,” he said.

The early-stage warning sign

The most worrying signal is at the early stage, where the next generation of growth companies is supposed to form.

Early-stage deal count fell to 195 in H1 2026, down 11 per cent year on year, 63 per cent below the H1 2022 peak and well short of the 353 deals recorded in H1 2024. If fewer startups are being funded today, the region could face a thinner pipeline of Series A, growth-stage and exit candidates in the years ahead.

Yet the amount of capital going into early-stage companies rose. Startups at this level raised US$1.72 billion, up 56.4 per cent year on year. Median pre-seed funding reached a series high of US$900,000, while median seed funding rose to US$3.7 million.

Investors, in other words, are writing larger cheques for a smaller group of companies. Founders who clear the bar may get more runway. Those outside investors’ preferred sectors or networks may find the door harder to open.

Series A remains the bottleneck. The median Series A round fell from US$11.6 million in H2 2025 to US$8 million, while the average dropped from US$17.6 million to US$12.9 million. Bigger seed rounds, therefore, do not necessarily mean more startups are graduating to institutional Series A financing.

Investors may instead be giving selected companies more time to prove product-market fit before facing a tougher priced round.

A broader but tougher capital stack

Another shift is the changing mix of capital providers. Sovereign funds, corporates, private equity investors and private credit providers are playing a larger role alongside traditional venture capital.

That can give founders access to deeper pools of money and strategic industry relationships. But these investors often assess risk differently from venture funds. They may care more about commercial traction, infrastructure value, strategic relevance or predictable cash flows than about high-growth narratives alone.

Debt is also becoming more common. Debt financing rose to 23 transactions from 17 a year earlier, with total value reaching US$1.36 billion. The implied average debt deal fell to US$59 million from about US$99 million in H1 2025, suggesting debt is being used by a wider set of companies rather than only a few large borrowers.

For Southeast Asian startups, this is both an opportunity and a constraint. The funding market is no longer only about persuading venture capitalists to bet on growth. Founders may need to assemble different types of capital for different stages of expansion.

Still waiting for a broad recovery

The macro backdrop is not weak. Developing Southeast Asia is expected to grow by about 4.6 per cent in 2026, while data-centre investment and technology exports are supporting markets such as Malaysia, Thailand and Vietnam.

But startup funding remains highly exposed to global conditions, including US interest rates, exit markets, currency risk and liquidity among limited partners. Until exits improve and more capital returns to regional venture funds, fundraising is likely to remain selective.

Also Read: 48 PE investors, US$3.96B deployed, and not a single IPO exit in five years. Something is broken.

The first half of 2026 shows that Southeast Asia can still attract large pools of capital when companies sit at the intersection of AI, infrastructure and global strategic demand. What it has not yet shown is a full recovery for the wider startup ecosystem.

For most founders, the funding winter has not ended. It has simply become more selective.

The post SEA startup funding jumps to US$7.25B, but most founders are still waiting appeared first on e27.

Indonesia’s SME lending ecosystem looks different in 2026 than it did five years ago. The peer-to-peer lending sector that produced KoinWorks, Investree, Modalku, Akseleran, and dozens of others has consolidated. The number of OJK-licensed P2P lenders has fallen sharply since 2022. The platforms that remain — KoinWorks among the most prominent — operate inside a tighter regulatory perimeter, with a more institutional funding mix, and with a credit posture that resembles traditional non-bank lending more than the original P2P model.

After fifteen years inside Indonesian risk functions, I have watched this transition with more interest than most parts of the industry. The SME credit gap the original P2P model was designed to address is still real — arguably larger now than it was in 2018.

What the P2P era built

The Indonesian P2P sector that emerged after OJK’s POJK 77/2016 did something the traditional banking system had not. It built credit access infrastructure for SMEs that conventional banks were not set up to serve — too small for commercial banking, too formal for microfinance, too thinly documented for traditional underwriting.

KoinWorks, founded by Benedicto Haryono and Willy Arifin in 2016, was one of the platforms that built deliberately for that segment from early on. Its emphasis on productive SME credit, on developing alternative data underwriting capability, and on maintaining a measured growth trajectory gave it a more durable position than many peers when the sector consolidated.

What changed

Three structural shifts have reshaped the sector.

Funding mix institutionalised. The original retail-investor-to-SME model has been steadily replaced by institutional funding — banks, asset managers, structured-credit vehicles. The platform’s role shifted from retail marketplace to credit origination intermediary.

Also Read: What I’m learning about the second wave of insurance digital transformation in Indonesia

Regulatory perimeter tightened. POJK 10/2022 and subsequent rules raised capital requirements, codified credit risk management expectations, and required clearer governance. Smaller platforms could not absorb the compliance cost. Consolidation followed.

Credit posture matured. The early sector underestimated default risk in the segments it served, partly because alternative data models were younger than the underwriting confidence they produced. Surviving platforms rebuilt credit policy around tighter limits, more conservative scoring, and active portfolio management.

Lessons learned

Six principles from this decade are worth carrying into the next chapter of Indonesian SME credit.

Discipline beats velocity. The platforms that survived grew slower than the market wanted them to. The ones that did not are mostly no longer licensed. Underwriting discipline is not a brake on growth — it is the condition for it.

Funding mix is survival, not treasury. A platform with multiple institutional funding lines has options. A platform with one funding line of any kind has a deadline.

Regulator engagement compounds. The platforms that spent time with OJK before the rules tightened got more flexibility when they tightened. Time-with-supervisors is the most under-priced asset in fintech.

Alternative data is a hypothesis, not a verdict. Models built on novel data require longer back-testing than founder optimism typically permits. Models should be continuously revisable, never declared proven.

Also Read: A 90-episode series in 3 weeks: How AI is speeding up Indonesia’s creative economy

The credit gap is durable, the model is not. Indonesia’s SME credit gap will exist as long as the banking system is structured the way it is. Founders who anchor on the gap rather than on the specific model will adapt faster.

Consolidation cycles repeat. The 2022-2024 P2P shake-out was not a one-off. The next category — embedded finance, vertical lenders, supply-chain credit — will go through its own version of this cycle inside five to seven years. The platforms that prepare for it instead of treating this round as the last one will still be operating after the next.

The macro stakes

KoinWorks and the platforms that came up with it gave Indonesia’s SME segment an underwriting infrastructure designed for them rather than adapted reluctantly from larger products. That contribution has not been fully absorbed by the formal banking system, and the gap remains for the next generation to address.

The lessons from the P2P era — discipline, diversification, alignment, humility about data, durability of the problem — are the foundation. The opportunity to build on them is still open. The institutions that build well in the next five years will be the ones that treat the previous five years as research, not as critique.

—

Editor’s note: e27 aims to foster thought leadership by publishing views from the community. You can also share your perspective by submitting an article, video, podcast, or infographic.

The views expressed in this article are those of the author and do not necessarily reflect the official policy or position of e27.

Join us on WhatsApp, Instagram, Facebook, X, and LinkedIn to stay connected.

The post The KoinWorks era: What Indonesia’s SME lending journey has taught the next generation appeared first on e27.

The global race to lead artificial intelligence is usually framed as a contest between countries: the US versus China, with Europe, India and a handful of others trying to carve out space. A new ranking argues that this view misses where much of the action is actually happening.

The Observer’s inaugural AI Cities Index, released today, maps AI capacity at the metropolitan level, scoring 57 cities across 36 countries. The report looks at three broad areas: investment, innovation and implementation. Put simply, it asks which cities are attracting money, producing breakthroughs, and building the institutions and infrastructure needed to use AI in the real world.

Also Read: Why Singapore, Indonesia, and Vietnam are losing the AI race they think they are winning

The answer still begins in the US. San Francisco and New York take the top two positions overall, underlining America’s continued strength in frontier AI research, venture capital and commercialisation. But the more striking finding is how quickly East and Southeast Asian cities are closing the gap.

Six of the world’s top 10 AI cities are in East and Southeast Asia: Seoul, Tokyo, Singapore, Shanghai, Shenzhen and Beijing. Together, they point to a shift in AI power away from a few US technology clusters and towards a more distributed network of Asian hubs, many of which lean on state coordination, industrial depth and public-sector adoption rather than venture funding alone.

The city as the new AI battleground

The Observer says the index builds on the Global AI Index, which has benchmarked national AI performance for seven years. Its city-level approach matters because AI ecosystems are rarely spread evenly across a country.

The US may be the world’s most influential AI market, but much of its cutting-edge activity is concentrated in and around San Francisco. China, by contrast, has three cities in the global top 10: Beijing, Shenzhen and Shanghai. South Korea’s AI strength is heavily concentrated in Seoul, while Singapore functions as both city and national AI platform.

This is particularly relevant for Southeast Asia, where AI adoption is unlikely to be driven by national scale alone. Singapore’s fifth-place ranking for implementation shows how a smaller market can punch above its weight when policy, talent, digital public infrastructure and enterprise adoption move in the same direction. For neighbouring markets such as Indonesia, Malaysia, Thailand, Vietnam and the Philippines, the question is not whether they can replicate Silicon Valley. It is whether they can build city-level clusters with enough talent, cloud capacity, industry demand and policy support to make AI useful beyond pilot projects.

Also Read: Korea’s startup ecosystem is training founders, not just funding them

The index defines implementation as the presence of institutions, systems and practitioners needed to operationalise AI across business, government, education and communities. That distinction is important. Building a powerful model is one thing. Putting AI into factories, hospitals, classrooms, banks and public services is another.

Seoul’s implementation edge

Seoul emerges as the most prominent non-US city in the ranking. While San Francisco tops the overall index, the South Korean capital ranks first globally for AI implementation.

That result reflects South Korea’s long-running investment in digital infrastructure, advanced manufacturing and semiconductors. Seoul is home to a dense cluster of semiconductor headquarters and related technology firms, including Samsung Electronics and SK Hynix, two of the world’s most important memory chipmakers. Both sit at the centre of the AI hardware boom, as demand for high-bandwidth memory and advanced chips rises with the growth of generative AI.

The city has also created the Seoul AI Innovation Committee to support small and medium-sized enterprises that cannot easily fund in-house AI talent or infrastructure. This is the kind of policy plumbing that rarely attracts the same attention as a new chatbot or chip launch, but it may prove more important in determining which economies actually benefit from AI.

For founders and operators in Southeast Asia, Seoul’s example is instructive. Many businesses in the region are not trying to train frontier models. They are trying to automate customer service, improve logistics, detect fraud, optimise energy use or equip workers with better tools. The winners may be cities that help ordinary companies adopt AI safely and affordably, not just those that host the biggest research labs.

Tokyo ranks third and Singapore fifth in the implementation table, while Shanghai, Shenzhen and Beijing also appear in the top 10. The report contrasts these cities with San Francisco and New York, where adoption is described as more fragmented and often still at pilot scale, despite deep private-sector innovation.

US still leads in capital and breakthroughs

None of this means the US is losing its AI lead. San Francisco remains the world’s strongest AI city overall, ranking highly across innovation, investment and implementation. New York also scores well across the board, helped by its deep capital markets, enterprise customer base, universities and growing AI startup scene.

On innovation, the report describes a more direct contest between Chinese cities and US technology hubs. San Francisco ranks first, followed by Beijing, New York and Shenzhen. This mirrors broader industry trends: the US continues to produce more frontier AI models, but Chinese labs and companies have narrowed the performance gap.

The Observer cites wider research showing that the performance gap between Chinese and American AI models has fallen to 2.7 per cent, down from as much as 31.6 per cent in 2023. The US still produced more frontier models in 2025, with 50 compared with China’s 30, but China’s count doubled year on year.

That narrowing gap matters for Asia’s startup ecosystem. If high-performing AI models become cheaper, more open and more widely available, the advantage may shift from those who own the models to those who know how to apply them in specific markets. Southeast Asian startups, often built around fragmented languages, regulations and customer behaviours, could benefit from this shift if they can localise AI effectively.

Europe struggles for space

Europe’s showing is comparatively modest. Only London and Paris make the global top 10. That reflects a familiar challenge: Europe has strong universities, research talent and regulatory influence, but has struggled to match the US in venture-backed scaling or East Asia in coordinated industrial deployment.

The ranking also suggests that state support alone does not guarantee implementation strength. Cities with prominent technology ambitions, including Tel Aviv and Dubai, do not make the top 10 for AI implementation.

Patricia Clarke, The Observer’s technology editor, said the index shows that AI power is being redrawn around cities rather than countries. “Some of tomorrow’s most important AI decisions may not be made in Washington or Beijing, but in Seoul, Shenzhen and a handful of other emerging hubs,” she said.

Also Read: A Southeast Asia AI adoption outlook vs alternative global hubs

For Southeast Asia, that is both a challenge and an opening. Singapore is already in the top tier for implementation, but the region’s larger markets still need deeper AI talent pools, stronger cloud and data infrastructure, clearer rules and more patient capital for applied AI. The cities that get those basics right may not dominate headlines like San Francisco, but they could determine how AI changes everyday business across the region.

The AI race is still led by the US. But if The Observer’s index is any guide, the next phase will be fought less by countries in the abstract and more by cities that can turn AI from promise into infrastructure.

The post Seoul tops global AI implementation as Singapore enters top tier appeared first on e27.

In 2022, Shein was valued by private investors at around US$100 billion, making it one of the world’s most valuable fashion companies.

This month, it went public in Hong Kong at a valuation of around US$27 billion — roughly 70 per cent below that 2022 peak. Its growth has also slowed: from 19 per cent in 2023 and 21 per cent in 2024, to eight per cent in 2025 and further again in 2026. Euronews questioned if it “signalled the death of ultra-fast fashion.”

There are plenty of reasons for that, from regulation and tariffs to investor scrutiny.

But there is also a bigger question worth asking: has the category that Shein helped build, evolved into something different?

How fast became ultra-fast

Fast fashion wasn’t new when Shein hit the market.

Zara and H&M had already changed the way people bought clothes, making trends available faster and at more accessible prices. The term fast fashion itself dates back to the early 1990s, when Zara’s model could take a garment from design to store in around 15 days.

Shein pushed that model much further.

Between July and December 2021, it was reportedly adding between 2,000 and 10,000 new products to its app every day. Small initial production runs allowed it to see what was selling, then quickly produce more of the winners.

The result was a clear category proposition: more choice, more often, at even lower prices. Shein came to define what became known as ultra-fast fashion.

But that category wasn’t created in isolation. It was enabled by a particular set of market conditions: global supply chains optimised for cost, access to low-cost production, inexpensive cross-border logistics and trade rules that made sending huge volumes of low-value parcels directly to consumers economically viable.

For years, those conditions supported the category, but recent conditions have changed.

The fast fashion fatigue

The US has removed the de minimis exemption that allowed low-value imports to enter without duties. Europe has introduced its own charge on low-value parcels.

France has gone one step further, with new legislation specifically targeting ultra-fast fashion, based partly on the volume of clothing a company puts onto the market and the cost of repairing a garment relative to its purchase price. The levy begins at different levels depending on the garment and can eventually reach €19.50 per item by 2030.

Also Read: Skate to where the puck will be: How category design gives you a breakaway

The characteristics that helped define the category are now becoming characteristics regulators use to identify and regulate it.

Environmental pressure adds a further layer.

Fast fashion’s model depends on high production volumes and short product cycles, with significant consequences for water, emissions and textile waste. Earth.Org estimates that 85 per cent of textiles end up in landfill each year, while synthetic textiles are also a major source of ocean microplastics.

These aren’t new problems — what’s changing is how difficult they are becoming for companies to treat as external to the business model. Regulation, trade policy and supply-chain scrutiny are increasingly turning them into questions of cost, compliance and competitiveness.

Then there is the customer.

People haven’t stopped wanting affordable clothes, but the consumer mindset that built the category is shifting.

Fast fashion solved a clear problem: more choice, more trends, at prices that made it possible to keep buying. Now, there are signs of fatigue with that cycle. Consumers are becoming more selective, with greater emphasis on quality, durability and value-per-wear over constant newness. The New York Times noted this shift in mindset as: ‘Buy Better, Buy less, Feel Smug About It.’

If consumers start valuing better over more, the advantage fast fashion was built around becomes less valuable. The opportunity is no longer to win harder at the existing category — but to re-think, and re-design the new category.

Categories aren’t fixed

We often talk about Category Design in the context of creating something new: identifying a problem the market hasn’t properly articulated, defining a different solution to it, and building a category around that new way of thinking.

But categories don’t stay still once they’re created.The customer changes. Technology changes. Regulation changes. New alternatives appear. What people value changes.

And eventually, the problem a category was designed to solve can start to look different too.

The mistake is assuming the category that made you successful will remain the right category simply because you lead it.

Netflix began with DVDs delivered by post, but the underlying customer need wasn’t DVDs. It was easier access to entertainment. As technology and behaviour changed, the category around that need changed too. Netflix was brilliant in recognising early that entertainment would one day be delivered digitally — something even its name anticipated. They saw where the category was heading and positioned themselves for it.

Also Read: Seizing opportunity when the competition blinks: Look for category and ecosystem openings

Category Design and ecosystem intelligence go hand in hand.

Categories are shaped by more than the companies competing within them. Customers, regulators, governments, analysts, investors and other stakeholders all influence what the market values, permits and expects. Tracking how those forces are moving in real-time is how you spot a category shift before it becomes obvious — and position yourself ahead of it.

What comes after ultra-fast fashion?

There are signs that Shein itself is already looking beyond the model that made it successful.

Following its IPO, it’s preparing for an acquisition-driven phase, targeting brands across different price points. Its first announced deal is Everlane, a brand associated with higher-quality basics, sustainability and supply-chain transparency — almost the opposite end of the fashion spectrum from Shein.

At the same time, Shein is expanding its marketplace and Xcelerator programme, giving other brands access to its manufacturing network, logistics and global customer base. It is using the infrastructure advantage it built in ultra-fast fashion to evolve beyond the current category it helped build.

Whether that represents an evolution of ultra-fast fashion or the beginning of something different is harder to say. If the category is changing, the answer isn’t simply to make fast fashion a little faster, cheaper or more efficient. It is to understand and design the new category to drive what consumers will value next.

Maybe value becomes less about the lowest possible purchase price and more about cost-per-wear.

Maybe resale becomes part of the original purchasing decision rather than something that happens afterwards.

Maybe consumers still want novelty, but expect brands to provide it with less waste.

Maybe supply-chain transparency itself becomes part of the value proposition, rather than something sitting behind the product.

Or maybe something else entirely becomes the basis on which the next generation of fashion companies competes.

The category isn’t clear yet.

And that’s exactly when Category Design matters.

—

Editor’s note: e27 aims to foster thought leadership by publishing views from the community. You can also share your perspective by submitting an article, video, podcast, or infographic.

The views expressed in this article are those of the author and do not necessarily reflect the official policy or position of e27.

Join us on WhatsApp, Instagram, Facebook, X, and LinkedIn to stay connected.

The post Is fast fashion losing its fit? appeared first on e27.

From project management and customer service to human resources (HR) and finance, businesses have access to more prebuilt software options than ever before.

However, buying prebuilt software that is designed to serve businesses across a range of industries is unlikely to accommodate your specific needs. That’s why many businesses are building custom tools, rather than adapting to generic software.

With three in four organisations embracing low-code development, building custom software has become just as accessible as buying generic tools.

Low-code and no-code platforms allow businesses to build custom tools whilst avoiding expensive software development projects and the need for coding knowledge. This supports faster prototyping and iteration, allowing businesses to test and refine tools based on user feedback.

This article explores four examples where building a custom tool makes more sense than buying prebuilt software.

• Your processes are unique or complex

Most prebuilt software is designed to serve many businesses at once, so its processes and workflows are built around common use cases. Buying this software may make sense for businesses with common, well-defined processes.

However, if your business has unique or complex processes, it is often a sign that building may be a better option than buying.

Low-code platforms allow businesses to create a solution that better reflects how they actually operate, rather than changing the way employees work to fit the limitations of prebuilt software.

• You want to integrate multiple systems

Businesses rarely rely on a single software, with teams using separate tools for project management, customer relationship management, finance, HR, reporting, and inventory.

Prebuilt software may offer some helpful integrations, but these are often limited. When your software doesn’t integrate properly, teams may have to manually transfer information between tools and spreadsheets, which can increase the risk of manual errors, disrupt productivity, and reduce employee satisfaction.

Custom software can be built specifically around your existing technology environment and how your business actually operates. For example, the warehouse team could upload inventory data that is automatically sent to the HR team responsible for orders.

Also Read: Why so many startups are cutting down on the number of tools they use

• You need more control over data 

Prebuilt software typically comes with security features determined by the provider. While these may be sufficient for many businesses, they may not provide the level of data and security control your business requires. This is particularly true if you are often handling sensitive information.

Custom software gives businesses greater control over how their data is collected, stored, accessed, and shared. You can incorporate data and security requirements directly into the software and create user permissions based on specific roles.

For example, a business could determine exactly which of their employees can access particular information, rather than relying on standard permissions provided by prebuilt software.

• Your business is adapting

If every business change requires purchasing another software, your technology stack can quickly become expensive and difficult to manage.

If you are expanding into new markets or changing the way your teams operate, consider building a custom tool that can adapt alongside your business.

Businesses can use low-code or no-code platforms to quickly add new fields, adjust workflows and processes, or update permissions to support additional users. This makes custom software a better long-term fit than prebuilt tools.

Also Read: The AI stack trap: Why more AI tools aren’t translating into more growth

Should you build instead of buy in 2026?

The goal is not to replace every prebuilt application with a custom-built alternative. Instead, it’s to have one custom-built software that accommodates your needs, processes, workflows, and business model.

Businesses should start by identifying where prebuilt tools are creating friction and consider whether a custom option would be better. This helps businesses avoid adding unnecessary tools to their technology stack and build solutions that better support how their teams actually work.

In most cases, businesses should consider building instead of buying when they have unique or complex processes, require seamless integration, need greater control over data and security, or are adapting and growing.

Low-code and no-code platforms ensure businesses no longer have to choose between adapting their processes to fit prebuilt software and commissioning a lengthy, expensive development project.

—

Editor’s note: e27 aims to foster thought leadership by publishing views from the community. You can also share your perspective by submitting an article, video, podcast, or infographic.

The views expressed in this article are those of the author and do not necessarily reflect the official policy or position of e27.

Join us on WhatsApp, Instagram, Facebook, X, and LinkedIn to stay connected.

The post Build vs buy: Why custom tools are winning in 2026 appeared first on e27.

Artificial intelligence has become one of the most exciting areas of healthcare innovation. From clinical documentation and diagnostics to patient engagement and remote monitoring, founders are finding new ways to bring AI into almost every part of the healthcare journey.

But for healthtech startups in Southeast Asia, having an impressive AI model may not be enough to build a successful company.

The harder and ultimately more valuable challenge is making technology work within real healthcare environments.

Healthcare is not simply a collection of datasets waiting for better algorithms. It is a complex network of patients, clinicians, hospitals, laboratories, pharmacies, insurers, medical devices, regulations, and existing software systems.

The startups that understand these connections may have a much better chance of moving from an interesting pilot to a product that healthcare organisations actually use.

The opportunity is real, but so is the complexity

Southeast Asia provides significant opportunities for digital health innovation.

Across the region, healthcare systems face growing demand, ageing populations, chronic disease burdens, uneven access to specialists, and differences between urban and rural healthcare.

Digital health can help address some of these challenges.

Telehealth can extend access beyond major cities. Remote monitoring can help clinicians follow patients outside hospitals. AI can support increasingly data-intensive clinical and operational work. Digital platforms can also make parts of the patient journey easier to navigate.

Yet Southeast Asia should not be treated as one homogeneous healthcare market.

Singapore, Indonesia, Malaysia, Vietnam, Thailand, and other regional markets differ in healthcare infrastructure, regulations, digital maturity, reimbursement models, and patient behaviour.

This means a healthtech product that succeeds in one market cannot always be introduced into another with only minor changes.

For founders, localisation needs to go much deeper than translating an interface.

The real problem may be workflow, not technology

Imagine an AI system that can identify a clinically relevant pattern in patient information.

Its accuracy may be impressive.

But what happens next?

Can the system access the right information from the hospital’s existing software? Does the result appear where a clinician already works? Can the clinician review or override the recommendation? Is the decision recorded properly? Can that information move to another system without being entered manually again?

Also Read: Vietnam’s healthtech boom has a talent problem nobody is talking about

If the answers are no, a sophisticated AI model can still become another disconnected tool.

This is why workflow integration deserves much more attention from healthtech founders.

Clinicians already work under considerable time pressure. Technology that adds another login, dashboard, or manual process may technically solve one problem while creating another operational burden.

Effective healthcare technology should fit naturally into how care is delivered.

That requires founders to understand the environment surrounding their product, not only the feature they are building.

Interoperability is becoming a product issue

For years, interoperability could largely be treated as an enterprise IT concern.

That distinction is becoming harder to maintain.

A digital health product may need to exchange information with electronic health records, laboratory systems, pharmacy platforms, medical devices, payer systems, or other healthcare applications.

When these systems cannot communicate effectively, the consequences become visible at the product level.

Users re-enter information. Clinicians switch between applications. Patient records become fragmented. Automation stops halfway through a workflow.

For a startup, interoperability therefore affects usability, adoption, and scalability.

Standards-based approaches can help, but supporting a technical standard is only part of the answer. Founders also need to understand how information moves through real healthcare processes and where their product belongs within those processes.

The question should shift from “Can our platform connect to another system?” to “Can information move through this care journey without unnecessary friction?”

AI needs healthcare context

The rapid improvement of generative AI has lowered the barrier to creating healthcare prototypes.

Building a demonstration is increasingly easy.

Building a dependable healthcare product remains difficult.

Healthcare AI operates in a field where incomplete context, inconsistent data, and inaccurate outputs can have consequences far beyond a poor user experience.

That makes human oversight especially important.

Rather than trying to remove healthcare professionals from the process, startups can design AI around them.

Also Read: Healthtech in South and Southeast Asia – Seeing beyond the “obvious”

A clinical documentation tool, for example, can prepare information for professional review instead of automatically treating generated content as final. A decision-support system can surface relevant information while leaving clinical judgement with the professional responsible for the patient.

This approach may appear less dramatic than the idea of autonomous healthcare, but it can create something much more valuable: trust.

Trust is also one of the hardest things for a young healthcare company to earn.

Infrastructure will separate pilots from scalable products

Many healthtech companies begin with a narrow use case, and that is often sensible.

Problems emerge when the underlying product is built only for that first use case.

A startup might initially serve one clinic, one hospital department, or one type of patient. Growth can later require supporting multiple organisations, new integrations, larger datasets, and different regulatory environments.

Architecture decisions made early can suddenly become business constraints.

This does not mean every startup needs enterprise-scale infrastructure from day one. Overengineering can be as damaging as underengineering.

Instead, founders should understand which technical decisions will be difficult to reverse.

Data architecture, security, interoperability, auditability, and the separation of core product capabilities from market-specific requirements deserve early consideration.

Scalability is not simply about handling more users. In healthcare, it also means handling greater organisational, technical, and clinical complexity.

The strongest healthtech founders will think beyond the feature

Southeast Asia does not lack healthcare problems worth solving, and it certainly does not lack entrepreneurial ambition.

The next phase of healthtech innovation, however, may reward companies that look beyond individual features.

Instead of asking only whether AI can perform a task, founders can ask whether that capability improves an actual healthcare workflow.

Instead of viewing integration as something to address after gaining customers, they can consider how the product will coexist with the systems healthcare organisations already depend on.

Instead of treating compliance, security, and governance as barriers to innovation, they can use them as foundations for building trust.

Instead of designing a product for an abstract Southeast Asian market, they can recognise that healthcare remains deeply local.

AI will undoubtedly influence the region’s healthcare future.

But the most successful companies may not be the ones with the most impressive AI demonstrations.

They may be the ones that solve the harder problem: making technology genuinely work within healthcare.

—

Editor’s note: e27 aims to foster thought leadership by publishing views from the community. You can also share your perspective by submitting an article, video, podcast, or infographic.

The views expressed in this article are those of the author and do not necessarily reflect the official policy or position of e27.

Join us on WhatsApp, Instagram, Facebook, X, and LinkedIn to stay connected.

The post Why Southeast Asia’s next healthtech winners will be built around healthcare workflows, not just AI appeared first on e27.

The next portfolio concentration problem may not come from investors staying too close to home, but from millions of investors becoming familiar with the same handful of global companies.

For decades, one of the most persistent puzzles in investing has been home bias.

Investors have traditionally tended to allocate disproportionately to companies and assets from their own countries, even when international diversification could give them access to a broader opportunity set. But something interesting may be happening among a new generation of digitally native investors in Southeast Asia. They are not necessarily staying close to home. They may be travelling too far in exactly the same direction.

In an early analysis of self-directed investor portfolios conducted while building Neufin, we encountered a pattern that challenged the conventional home-bias assumption. Across a sample of self-selected digital-investor portfolios, US-listed global companies repeatedly dominated the holdings.

Apple, Microsoft and JPMorgan each appeared in more than 30 per cent of the sampled portfolios. Even more strikingly, no SGX-listed security appeared among the 30 most frequently observed holdings in the sample. The dataset is small and self-selected. It should not be interpreted as representative of investors across Singapore or Southeast Asia. But the pattern raises a much bigger question. What if the digital era is not eliminating home bias, but reversing it?

When familiarity is no longer geographical

Traditional home bias is intuitive. People understand the companies they encounter around them. They recognise their local banks, telecom operators, property groups and retailers. Geography creates familiarity, and familiarity influences investment decisions. Digital investing changes what familiar means.

A Singaporean, Malaysian, Indonesian or Vietnamese investor may interact with Apple products every day, search through Google, watch Netflix, use Microsoft software, read about Nvidia and Amazon, and encounter commentary on these companies continuously across financial media and social platforms. These businesses may be headquartered thousands of kilometres away, but psychologically they can feel closer than a locally listed industrial or property company.

The investor’s geographical home has not moved. Their information home has. That creates what I would describe as reverse home bias: a tendency for digitally connected investors outside the US to disproportionately gravitate towards a relatively small universe of globally familiar securities. This is not simply international diversification. It is familiarity bias operating across borders.

The diversification illusion

There is an important distinction between owning international companies and being genuinely diversified. Consider a hypothetical portfolio containing Apple, Microsoft, Nvidia, Amazon, Alphabet, Meta and Tesla. At first glance, it looks sophisticated and global. These are enormous businesses serving customers across many countries and industries. Yet at the portfolio level, the investor may still have significant common exposures. Several holdings can respond simultaneously to US interest-rate expectations, technology valuations, movements in the US dollar, American economic policy, changes in risk appetite or a broad sell-off in growth equities.

Also Read: India’s IPO boom is rewriting the exit playbook for global investors

The logos are different. The underlying risk drivers may be much less different than they appear. That matters because conventional portfolio interfaces tend to emphasise what an investor owns: ticker symbols, sectors, countries, gains and losses. They are less effective at showing why those positions may move together. A portfolio can therefore appear diversified at the security level while remaining concentrated at the behavioural, factor or narrative level.

The algorithmic layer could make this stronger

There is another reason reverse home bias deserves attention now. Investment discovery is becoming increasingly algorithmic. A decade ago, an investor might have discovered companies through a broker, newspaper, research report or financial adviser. Today, discovery can begin with a YouTube video, TikTok clip, Reddit discussion, financial app notification, search engine or increasingly an AI assistant.

This changes the mechanics of familiarity. Algorithms naturally amplify information that already has high visibility, engagement and availability. Large global companies have enormous digital footprints: thousands of articles, earnings transcripts, analyst reports, videos, discussions and historical references. Generative AI introduces another layer.

Ask an AI system for examples of leading technology companies, innovative businesses or investment themes and globally documented companies have an obvious informational advantage. They exist abundantly within the digital knowledge environment from which answers are constructed.

This does not mean AI will automatically recommend American equities, nor that investors will blindly follow AI-generated information. But it creates a question worth examining:

Could AI-mediated financial discovery make the world’s most information-rich companies even more cognitively dominant? If so, the next generation of investor bias may be shaped as much by information architecture as by geography.

Why this matters for Southeast Asia’s fintech ecosystem

For wealthtech platforms, advisers and financial institutions, this is more than an interesting behavioural-finance observation. Most suitability and portfolio-review processes are designed around relatively visible risks.

How much equity exposure does the client have? How concentrated is the portfolio? What is the client’s risk tolerance? What sectors and geographies are represented? Has the asset allocation drifted? Those remain important questions.

But digital investing may require additional ones. How many holdings are ultimately driven by similar macroeconomic factors? How much of the portfolio reflects the same investment narrative? Does apparent geographical diversification conceal currency or factor concentration? Are holdings becoming more concentrated because of repeated digital exposure? Has the client’s portfolio gradually moved away from their stated risk profile even though each individual trade seemed reasonable at the time? And eventually, one particularly difficult question:

Is an investor choosing an asset because it fits the portfolio, or because the asset has become exceptionally visible to them? These are difficult questions for a traditional portfolio dashboard to answer. They are increasingly important questions for an AI-enabled financial system.

Also Read: How to pitch Southeast Asia’s investors: A founder’s guide

Reverse home bias is not an argument for buying local

There is an important distinction here.

Reverse home bias should not become an argument that Southeast Asian investors ought to own more domestic securities simply because they are domestic. That would replace one familiarity bias with another. US markets offer extraordinary businesses, deep liquidity and access to industries that may be difficult to obtain through local exchanges. International exposure can be an important part of portfolio construction.

The issue is not whether Apple is preferable to a Singapore-listed company, or whether an investor should allocate a particular percentage to one market. The issue is whether familiarity is being mistaken for diversification. A rational global portfolio and a globally familiar portfolio are not necessarily the same thing.

We may need a new definition of portfolio intelligence

The investment industry has become exceptionally good at describing portfolios. The next challenge is understanding the context behind them. A future portfolio-risk system may need to evaluate more than securities, sectors and historical volatility. It may also need to understand behavioural concentration, correlated narratives, suitability drift, attention patterns and the sequence of decisions that created the portfolio.

That becomes particularly important as AI begins participating more deeply in financial research, recommendations and eventually financial actions. The critical question will no longer be simply: “What does this investor own?” It will increasingly become: “Why does this investor own these assets, what common risks sit underneath them, and is the portfolio still consistent with what the investor is trying to achieve?”

Our initial sample of 82 portfolios is nowhere near sufficient to declare reverse home bias a regional phenomenon.

But it is enough to make the hypothesis worth testing. If larger datasets reveal the same pattern, Southeast Asia may offer an early view of a significant change in investor behaviour: a world in which capital becomes geographically global while investor attention becomes increasingly concentrated.

The old home bias was created by proximity. The new one may be created by visibility. And in an AI-mediated financial world, visibility could become one of the most important investment biases we learn to measure.

—

The post Reverse home bias: Why Southeast Asia’s digital investors may be diversifying in the wrong direction appeared first on e27.

For decades, the semiconductor industry has relied on one basic bargain: make silicon transistors smaller, and chips become faster, cheaper and more power-efficient. That bargain is getting harder to keep.

As AI models grow larger and data centres consume more electricity, chipmakers are running into a materials problem. Silicon, the foundation of modern computing, can only be shrunk so far before electrons begin to leak through transistor channels, wasting power and limiting performance gains.

Singapore-based Nexstrom believes the next step will not come from squeezing more out of silicon, but from replacing parts of the transistor with atomically thin semiconductor materials.

Also Read: NASA-linked, MIT-trained founders’ nSWX raises US$2M for AI chip packaging

The company has secured US$12 million in seed funding led by Xora Innovation, with participation from Foothill Ventures and SEEDS, an arm of SG Growth Capital. The round brings Nexstrom’s total capital raised to US$15 million, including US$3 million in non-dilutive funding.

Nexstrom said the capital will be used to commercialise its wafer-scale platform for two-dimensional (2D) semiconductor materials. Its near-term target is ambitious: to develop what it describes as the industry’s first 12-inch single-crystal 2D semiconductor wafer growth platform using production-ready manufacturing tools.

That matters because 12-inch, or 300 mm, wafers are the standard used by advanced chip foundries. Many promising semiconductor materials have performed well in laboratories but failed to make the jump to large, uniform wafers that can survive commercial manufacturing.

Why 2D materials matter

2D semiconductors are materials only a few atoms thick. One of the best-known examples is molybdenum disulphide, or MoS₂, a compound that has attracted research attention because it can help control electron flow at extremely small dimensions.

In simple terms, thinner channels give chipmakers better control over the movement of electrons. That could reduce leakage, lower power consumption and support further transistor scaling at a time when silicon is becoming harder to push.

This is especially relevant for AI and high-performance computing, where the economics of performance are increasingly tied to energy efficiency. Training and running large AI models already require vast amounts of computing power, and every incremental gain in chip efficiency can translate into meaningful savings for hyperscale data centres.

But the obstacle has never been scientific promise alone. The bigger question is whether 2D materials can be produced at the scale, consistency and cost that advanced foundries demand.

Also Read: Synopsys, A*STAR team up to tackle AI chip packaging challenges

“Silicon has fuelled decades of computing innovation, but the industry now needs a new materials platform to continue scaling performance,” said Dr Lance Li, Nexstrom’s co-founder and Chief Scientist. “For years, the challenge has not been demonstrating the promise of 2D materials, but manufacturing them at the scale and quality advanced foundries require.”

From research to foundry floor

Nexstrom was founded in 2024 and incubated through Xora Innovation’s venture-building model, which focuses on turning deeptech research into companies that can address industrial markets.

Its platform combines proprietary chemical vapour deposition hardware, process technology and wafer-scale 2D material growth. Chemical vapour deposition, or CVD, is a manufacturing process used to deposit thin films of material onto wafers.

Nexstrom’s approach is designed to fit into existing foundry workflows rather than requiring chipmakers to rebuild their manufacturing lines from scratch.

That compatibility will be important. In semiconductors, even promising materials face long adoption cycles because foundries are highly conservative environments. Any new process must deliver uniformity, repeatability and yield, while fitting into an industry already built around expensive equipment and tightly controlled production steps.

Nexstrom says its technology is aimed at continuous, single-crystal 2D material growth across 12-inch wafers. “Single-crystal” refers to material with a uniform atomic structure, which is important because defects and grain boundaries can affect electronic performance. In commercial chipmaking, uniformity across the wafer is just as important as performance in a single device.

The company is working with industry partners to validate its technology within existing chip manufacturing workflows. It did not name those partners.

A Singapore bet on upstream semiconductor technology

For Southeast Asia, Nexstrom’s financing lands at an interesting moment. The region has long been part of the global semiconductor supply chain, particularly in assembly, testing, packaging and equipment services. Singapore, Malaysia, Vietnam and the Philippines all play important roles in chip production, though most cutting-edge logic manufacturing remains concentrated in Taiwan, South Korea and the US.

Singapore has been trying to move further upstream, building on its base of wafer fabrication, precision engineering and research talent. The city-state already hosts operations from major semiconductor companies and has pushed deep tech as a strategic priority through public funding, university research and state-linked investment vehicles.

Nexstrom fits into that broader shift. Rather than building another application-layer AI company, it is attempting to address a bottleneck much closer to the physical foundations of computing. If successful, such technology would be relevant not only to AI chips, but also to data centres, advanced processors and future low-power electronics.

Still, the road from seed-stage materials company to semiconductor supplier is unusually long. Deep-tech hardware startups face lengthy validation cycles, high capital requirements and demanding customers. Unlike software companies, they cannot iterate through code alone; they must prove performance in physical systems, often over many years.

The competitive field

Nexstrom is entering a race that includes some of the world’s most sophisticated chipmakers, equipment companies and research institutes. Major players such as TSMC, Samsung, Intel and imec have explored 2D materials as possible candidates for future transistor channels, while universities and specialist materials companies are also working on graphene, MoS₂ and other post-silicon approaches.

Its challenge, therefore, is not merely to show a better material, but to build a platform that foundries can realistically adopt. That may be where a focused startup has room to compete: by solving one narrow but critical manufacturing bottleneck rather than trying to build an entire chip ecosystem around the technology.

Also Read: Southeast Asia’s chip-hub ambition is colliding with its chip-smuggling problem

Nexstrom’s technical foundation is led by Li, a recognised researcher in 2D materials and a Clarivate Highly Cited Researcher since 2018, a distinction covering roughly the top 0.1 per cent of researchers globally by citation influence. He has worked on single-crystal MoS₂ growth since 2012 and later led corporate research at TSMC on post-silicon electronics.

The company is also supported by advisors including Dr Sundar Ramamurthy, Dr Philip Wong, Dr Aaron Thean and Dr John Langan.

“The question is no longer whether 2D semiconductors matter. They are already on the technology roadmap of major semiconductor companies. The challenge is making them manufacturable,” said Wong, Board Advisor at Nexstrom and Inez Kerr Bell Professor at Stanford University.

That line captures both the opportunity and the risk. The semiconductor industry knows it needs new materials to keep scaling performance. But knowing what comes next is different from manufacturing it at commercial scale.

With its new funding, Nexstrom will expand platform development, deepen collaborations with foundries and grow its engineering and leadership teams. For Singapore’s deep-tech ecosystem, the company will be one to watch: not because 2D semiconductors are guaranteed to win, but because the next era of computing may depend on companies willing to work at the atomic edge of what silicon can no longer do.

The post Nexstrom lands US$12M to bring 2D semiconductors to 12-inch wafers appeared first on e27.