Day: September 10, 2026

Speculators rapidly adjust portfolios in response to shifting interest rate expectations and escalating global conflicts. The leading cryptocurrency recently experienced a notable decline while major United States stock indices suffered significant losses. This dual downturn highlights a broader risk-off strategy among individuals anticipating tighter monetary policy and higher energy costs. These factors create a highly volatile environment demanding careful analysis of underlying metrics rather than superficial valuation movements. We must evaluate specific numbers driving these asset classes to understand true directional momentum.

The top digital token currently trades at US$78,510.84, down 0.77 per cent over the last 24 hours. This instrument slightly underperforms a relatively flat broader financial landscape. Such divergence indicates crypto-specific macro positioning drives current valuation action rather than general equity trends. We see a weak correlation between traditional safe havens and stocks during this specific window.

The S&P 500 moved down just 0.06 per cent while Gold gained 0.29 per cent over the same period. Participants currently treat the primary decentralised network strictly as a rate-sensitive risk instrument. They reduce exposure to non-yielding speculative tokens as risk-free government bond yields climb. This behaviour confirms that digital currency ecosystems operate with unique internal dynamics when confronted with shifting monetary landscapes.

The dominant driver behind this crypto ecosystem shift involves changing rate expectations. Last week saw a stronger-than-expected United States jobs report, which added 162,000 positions to the economy. This robust employment data immediately increased odds for a Federal Reserve rate hike at the upcoming September 15 to 16 meeting. Higher government bond yields directly reduce the relative appeal of speculative investments.

Buyers now heavily price in a higher probability of tighter monetary policy ahead of critical inflation figures. The financial world eagerly awaits the United States Consumer Price Index report scheduled for Friday, September 11. This upcoming inflation print will either solidify or soften current expectations for central bank hikes. Market participants remain highly sensitive to economic metrics that might influence monetary authority decisions.

Also Read: Why a strong jobs report hit Bitcoin and Ethereum harder than the stock market

Valuation drops in the digital asset space frequently trigger severe mechanical selling. The recent cryptocurrency decline initiated a massive leverage unwinding event across the digital landscape. Exchanges recorded US$264 million in total liquidations over the last 24 hours. Positions tied to the largest blockchain accounted for US$73.33 million of this total. This liquidation volume represents a 76.73 per cent increase from the previous day.

Approximately 90 per cent of these forced sales involved long positions. This statistic indicates a complete flush of overleveraged bullish bets. Forced selling creates a dangerous feedback loop that exacerbates downward valuation momentum. Analysts must watch for stabilisation in open interest and funding rates to confirm that this leverage flush has finally run its course across major platforms.

Traditional equity venues also reflect deep participant concern regarding the broader economic outlook. Wall Street closed lower on Tuesday, September 8, 2026. Major indices surrendered substantial ground as individuals digested negative news regarding global energy supplies. The Dow Jones Industrial Average fell 628.18 points or 1.2 per cent to close at 52,786.07. The S&P 500 dropped 45.08 points or 0.6 per cent to finish at 7,673.52.

The Nasdaq Composite slipped 85.58 points or 0.3 per cent to end the session at 26,421.41. Small-cap stocks also retreated as the Russell 2000 index lost 15.44 points, or 0.5 per cent, to settle at 2,960.20. These broad declines demonstrate that equity buyers share the exact same risk-off sentiment currently gripping the digital asset space and global commodity venues.

Sector performance on Wall Street clearly illustrates the specific fears driving this equity sell-off. The Energy, Utilities, and Real Estate sectors managed to close higher despite the broader index’s decline. Conversely, Health Care, Financials, and Materials severely lagged the wider financial landscape.

The Dow Jones Industrial Average took a particularly hard hit due to a sharp pullback in healthcare. The primary catalyst for this sector rotation involves rapidly escalating geopolitical tensions involving Iran. These conflicts directly threaten global energy infrastructure and disrupt regional supply chains. Crude oil markets reacted violently to these disruptions. Brent crude briefly approached US$99.50 a barrel as attacks on regional energy facilities spooked commodity buyers. Oil eventually settled in the green as participants priced in a sustained period of elevated energy costs.

Also Read: From US$79,300 to US$82,400: Mapping the narrow corridor that decides Bitcoin’s September

Rising energy costs directly renew inflation worries among institutional and retail buyers. Crude oil nearing US$100 a barrel introduces a massive variable into future inflation calculations. Higher fuel and transportation costs inevitably filter down to consumer goods and services. This dynamic severely complicates the Federal Reserve’s mandate to maintain price stability.

The combination of strong jobs data and surging oil prices creates a perfect storm for persistent inflation. Participants now fear that the central bank might adopt an even more aggressive stance to combat these rising prices. This reality explains why both digital assets and traditional equities sold off simultaneously. Individuals simply lack the appetite to hold risk instruments when the cost of capital threatens to rise significantly soon.

The immediate outlook for the leading cryptocurrency remains sideways to bearish until the ecosystem digests upcoming inflation data. A sustained hold above US$78,000 could stabilise the asset and attract cautious buyers. A daily close below this crucial threshold would likely trigger a test of lower supports. Analysts currently target the US$76,000-US$77,600 support zone if bearish momentum continues.

Also Read: Bitcoin just broke US$81,000: The real reason is not what you think

Conversely, a cooler inflation print could allow a rebound toward the US$80,000 mark. Speculators should note that firm resistance awaits near the US$81,000 to US$82,000 range. Institutional exchange-traded fund demand continues to provide a structural bid for the asset. Macroeconomic fears completely overshadow this underlying institutional demand during the current trading week as participants await concrete economic numbers.

The combination of hawkish central bank repricing and a leveraged long squeeze has definitively tipped short-term momentum downward. The path of least resistance remains cautiously lower until the ecosystem receives concrete macroeconomic confirmation. Friday, September 11, stands out as the most critical date for near-term price discovery.

The reaction at the US$78,000 support will determine whether the financial landscape experiences a healthy pullback or a deeper correction. At the time of writing this analysis, Asian exchanges have not opened for the trading session. I eagerly anticipate analysing the Asian exchange reaction when trading begins. The opening bell in Asia will likely provide crucial clues regarding global sentiment and set the tone for the week.

—

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 Asia has not opened yet: What will the first bell reveal about Bitcoin and oil? appeared first on e27.

Thailand’s mobility story is often told through the lens of electric vehicles, new car factories and government incentives. But at HERE Directions Bangkok 2026, the more important question was not simply what people will drive. It was how vehicles, roads, cities and public agencies will share enough intelligence to make movement safer and more efficient.

Hosted by location data company HERE Technologies at Park Hyatt Bangkok, with Amazon Web Services as co-host, the event brought together government representatives, automakers, technology firms and mobility specialists to examine the next phase of Thailand’s transport evolution. The discussion ranged from road safety and electrification to AI-assisted driving and connected urban data systems.

Also Read: Southeast Asia’s EV startups draw US$622M as clean mobility shifts from pitch to pilot

The timing matters. Thailand is already Southeast Asia’s most important automotive manufacturing base and has set an ambition for zero-emission vehicles to account for 30 per cent of total vehicle production by 2030. At the same time, Bangkok’s congestion, Thailand’s high road fatality rate and the country’s heavy dependence on motorcycles show that the mobility transition cannot be solved by swapping petrol engines for batteries alone.

Road safety is still the hardest problem

One of the clearest themes from the event was that road safety remains Thailand’s most urgent mobility challenge.

Motorcycles account for nearly half of registered vehicles in the country, making two-wheeler safety central to any national transport strategy. For millions of Thais, motorcycles are not recreational vehicles; they are the default option for commuting, food delivery, informal logistics and last-mile transport. That makes the risks harder to manage and the policy response more complex.

Location intelligence has an obvious role here. Accident hotspot alerts, road condition notifications, safer route suggestions and live traffic updates can help drivers and riders make better decisions before danger becomes unavoidable. For fleet operators, the same data can shape driver coaching, route planning and insurance risk models.

“Thailand is entering a new era of mobility where real-time decisions matter more than ever. Electrification, AI-powered driving experiences and rising expectations around road safety are transforming how people and goods move,” said Deon Newman, Senior Vice President and General Manager for Asia Pacific at HERE Technologies. “As vehicles become more connected and software-defined, location intelligence is becoming the critical foundation that helps drivers, businesses and cities make safer, smarter and more informed decisions in real time.”

The point is especially relevant in Southeast Asia, where urban transport systems are highly mixed. Cars, buses, motorcycles, tuk-tuks, delivery riders and pedestrians often share the same road space. In that environment, maps cannot be static digital replicas of roads. They need to capture changing conditions, risk patterns and local driving behaviour.

EV adoption needs more than chargers

Electrification was another major focus, but speakers treated it as part of a wider mobility shift rather than a standalone vehicle trend.

Thailand’s EV market has been expanding, supported by government incentives and investment from global and Chinese automakers. Yet the transition includes more than battery electric cars. Hybrids, plug-in hybrids, hydrogen fuel-cell vehicles, commercial fleets, electric buses and two-wheelers will all be part of the mix.

This creates a new planning burden. Drivers need to know not only where a charging station is, but whether it is available, compatible, reliable and reachable based on battery level, traffic and terrain. Fleet operators need to plan routes around charging windows and delivery schedules. Cities need to understand where infrastructure gaps are emerging.

That is where location data becomes operational rather than merely navigational. A map that can combine road networks, charging locations, energy consumption patterns and live traffic can help reduce range anxiety and improve vehicle utilisation. For logistics companies, even small gains in routing efficiency can translate into lower costs across thousands of trips.

AI turns maps into decision systems

The event also looked at how artificial intelligence is changing the role of in-vehicle navigation. Advanced driver assistance systems, or ADAS, and Navigation on Autopilot are pushing maps beyond turn-by-turn directions.

For these systems to work safely, vehicles need to understand road context: lane structures, speed restrictions, intersections, curves, construction zones and hazards ahead. AI can help interpret this environment, but it still depends on reliable underlying map and location data.

Also Read: SLEEK EV’s US$8.5M Series A funding signals a more mature EV playbook

This is where the industry is moving towards what automakers often call software-defined vehicles. In simple terms, more of the vehicle’s functions are managed and improved through software rather than fixed hardware alone. Navigation, safety alerts, driver assistance and infotainment are increasingly connected.

At HERE Directions Bangkok, neueHCT demonstrated several AI-powered driving technologies, including HCT Astra, an assisted driving platform; HCT Luna, a smart camera system; and HCT Orbis, a rider assistance system for two-wheelers. The inclusion of two-wheeler technology is notable in Thailand and the wider region, where mobility innovation often needs to start with motorcycles rather than premium cars.

Smarter cities need shared data

Beyond vehicles, the event returned repeatedly to the importance of connected data ecosystems. Smart city projects often struggle because information sits in separate systems across government agencies, transport operators, emergency services and private mobility companies.

Dr Passakon Prathombutr, Vice Chairman of iTIC and Special Expert at Thailand’s Digital Economy Promotion Agency, argued that the value of data increases when different layers can be combined. Accident data, GPS traces, road context and infrastructure information can reveal patterns that would be invisible in isolation.

“Building smarter and more sustainable cities requires more than technology. It requires the ability to connect data across agencies, infrastructure and mobility ecosystems,” he said. “When data can move seamlessly between stakeholders, this can add higher value to cities that can gain deeper insights, improve decision-making and deliver more efficient, safer and citizen-centric mobility services.”

For Bangkok, this is not an abstract ambition. The city’s transport challenges are shaped by density, legacy infrastructure, flooding risks, delivery growth and fragmented public transport options. Better data sharing could support traffic management, emergency response, public transport planning and road safety interventions.

A competitive mapping race

HERE Technologies is not alone in chasing this opportunity. Globally, it competes with Google Maps Platform, TomTom and Mapbox across mapping, location services and automotive navigation. In Southeast Asia, Grab has also built mapping capabilities to support ride-hailing, deliveries and logistics, while Waze remains influential in crowdsourced traffic information.

The competitive landscape reflects a broader shift: mapping is no longer just a consumer app category. It is becoming infrastructure for autonomous driving, urban planning, insurance, logistics, advertising, EV charging and public safety. For Thailand, that means the winning solutions will need strong local data, partnerships with public agencies and the ability to work across messy real-world conditions.

Also Read: Thailand’s AI startup push gets OpenAI backing through new public-private accelerator

The lesson from HERE Directions Bangkok 2026 is that Thailand’s mobility future will depend less on any single technology than on how well different systems talk to one another. EVs, AI-assisted driving and smart city platforms may capture the headlines, but their impact will be limited if roads, vehicles and institutions continue operating on incomplete information.

In Southeast Asia, where mobility is crowded, informal and fast-changing, the next breakthrough may not look like a futuristic car. It may be a better decision made a few seconds earlier — by a driver, a fleet manager, a traffic controller or a city planner.

The post Thailand’s mobility future will be decided by data, not just vehicles appeared first on e27.

For years, the semiconductor race was largely about making transistors smaller. That contest is far from over, but the AI boom has shifted part of the battleground elsewhere: how multiple chips are assembled, connected and kept reliable inside a single package.

That is the problem Synopsys and Singapore’s Agency for Science, Technology and Research (A*STAR) are now trying to address.

Also Read: The factories are coming. Southeast Asia’s real race is to build what surrounds them

The US-based chip design software company and Singapore’s national research agency have signed a memorandum of understanding to jointly develop advanced semiconductor-packaging and simulation technologies for artificial intelligence and high-performance computing.

The collaboration will focus on advanced packaging and chiplet-based designs. Chiplets are smaller specialised chips that can be combined in one package to function like a larger, more powerful system. Instead of relying on one monolithic chip to do everything, companies can mix and match computing, memory, networking and other functions in a more modular way.

This approach is increasingly important for AI and high-performance computing, where systems need far more processing power, memory bandwidth and energy efficiency than traditional chip designs can easily deliver. But it also creates new engineering challenges. When several chips are packed tightly together, heat, stress, warping and material behaviour become harder to predict.

Synopsys said digital modelling can help assess how a package is likely to perform before companies spend time and money on physical prototypes. In an industry where development cycles are long and fabrication mistakes are expensive, being able to simulate reliability early can make a meaningful difference.

Why packaging now matters as much as design

The collaboration will run through the ASTAR IME-Ansys Joint Innovation Consortium for Semiconductor Excellence, involving ASTAR’s Institute of Microelectronics and Ansys, which is now part of Synopsys. The consortium will act as a platform for companies and researchers to conduct joint research on advanced System-in-Package technologies.

System-in-Package, or SiP, refers to the integration of multiple chips or components inside one package. It is already used in areas such as smartphones, automotive electronics and wearables, but AI computing is pushing the technology to far greater levels of complexity.

The initial phase of the consortium’s work will focus on mechanical design, modelling and analysis. The partners aim to examine issues including package and wafer warping, thermo-mechanical stress, solder-joint reliability and moisture-induced failures in multi-chiplet designs.

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

These may sound like back-end engineering details, but they are central to whether future AI systems can be manufactured at scale. A high-performance chip package may fail if it bends during production, develops microscopic cracks under heat, or suffers from unreliable solder joints over time. For AI data centres, autonomous systems, advanced manufacturing and next-generation consumer electronics, reliability is not optional.

ASTAR IME will lead the consortium’s research, drawing on its semiconductor packaging capabilities and research platforms. Synopsys will provide trial licences to its Ansys simulation software to ASTAR IME and up to 10 member companies. The consortium also plans to explore projects with universities and offer technical training.

That training element is particularly relevant for Singapore and the wider region. Semiconductor ecosystems are not built only on fabs and equipment; they also require engineers who understand materials, design, thermal behaviour, electronics, manufacturing constraints and simulation tools.

Singapore’s semiconductor bet

Singapore has long played an outsized role in the global semiconductor supply chain. It is home to wafer fabrication, assembly and test operations, equipment suppliers, materials companies and regional headquarters for multinational chip firms. While it does not compete with Taiwan or South Korea in leading-edge logic manufacturing, it has positioned itself as a serious hub for specialty chips, advanced packaging, research and manufacturing services.

That positioning matters as the global chip industry becomes more geopolitically fragmented. The United States, China, Europe, Japan, South Korea and Taiwan are all investing heavily in semiconductor capabilities, driven by AI demand and concerns over supply chain resilience. Southeast Asia, meanwhile, has become more important as companies diversify manufacturing footprints and look for politically stable, technically capable locations.

Malaysia is already a major assembly and testing hub, particularly in Penang. Vietnam is attracting interest in chip design and back-end manufacturing. Thailand and the Philippines have existing electronics manufacturing bases. Singapore’s advantage lies in its research institutions, talent base, intellectual-property protections and proximity to both global companies and regional manufacturing networks.

The Synopsys-A*STAR partnership fits neatly into that strategy. Rather than trying to win every part of the chip supply chain, Singapore has been focusing on areas where deep engineering, industry collaboration and applied research can create defensible value.

Terence Gan, Executive Director of A*STAR IME, said advanced packaging is becoming a critical differentiator in chip innovation as systems grow more complex and AI-driven. He added that the collaboration can reinforce Singapore’s position as a semiconductor innovation hub.

Simulation becomes a strategic layer

For Synopsys, the tie-up also reflects how electronic design automation companies are expanding beyond traditional chip design software. The company’s acquisition of Ansys brought simulation capabilities closer to chip and system design, at a time when the boundaries between semiconductor design, packaging and system-level engineering are blurring.

In AI hardware, performance is no longer determined only by the processor. Memory access, interconnects, power delivery, cooling and package architecture all shape the final system. This makes simulation more strategic. Companies need to understand how a design will behave electrically, mechanically and thermally before it reaches production.

Sukhwan Moon, Synopsys Vice President of Asia-Pacific Sales, said the collaboration can support local companies in reliability, performance and scalability, while helping speed time-to-market for AI and high-performance computing technologies.

That time-to-market pressure is intense. AI infrastructure demand has created a rush for more powerful chips, faster networking and more efficient computing systems. Cloud providers, hyperscalers, chip startups and electronics manufacturers are all trying to move quickly, but the hardware cycle remains unforgiving. Mistakes in design or packaging can delay products by months.

The competitive landscape

Synopsys operates in a highly concentrated but fiercely competitive market. Its main global rivals include Cadence Design Systems and Siemens EDA, which also provide chip design, verification and electronic design automation tools. In simulation and engineering software, the enlarged Synopsys now overlaps with companies such as Keysight Technologies, Altair and Dassault Systèmes in certain areas, depending on the application.

Also Read: AI demand lifts Malaysia’s chip sector, but not every player wins

In advanced packaging, competition is not limited to software. Foundries, outsourced semiconductor assembly and test players, and integrated device manufacturers are all building capabilities around chiplets, 2.5D and 3D packaging. TSMC, Intel, Samsung, ASE, Amkor and JCET are among the companies shaping this market globally. For Southeast Asia, this creates both opportunity and pressure: the region can capture more value, but only if it moves beyond low-cost manufacturing into higher-value engineering and research.

A small MoU in a much larger race

The MoU between Synopsys and A*STAR is not a chip factory announcement, nor does it come with the headline-grabbing capital expenditure often associated with semiconductor projects. Its importance lies elsewhere.

Advanced packaging is becoming one of the key ways the chip industry keeps improving performance as traditional scaling becomes harder and more expensive. For AI, where demand for computing power continues to surge, the ability to design reliable multi-chip systems could determine which companies and countries capture the next wave of value.

For Singapore, the partnership strengthens a role it has been cultivating for years: a neutral, research-driven and industry-connected node in the global semiconductor network. For Southeast Asia, it is another sign that the region’s chip opportunity is widening beyond assembly lines.

The future of AI hardware will not be decided only in data centres or wafer fabs. It will also be shaped in the less visible world of packaging labs, simulation platforms and reliability testing — precisely where this collaboration intends to operate.

The post Synopsys, A*STAR team up to tackle AI chip packaging challenges appeared first on e27.

Every hyperscaler courting Southeast Asia now performs the same reassurance ritual. Ask Microsoft, Google, or AWS about the environmental cost of the data centres they are racing to build across the region, and the answer arrives pre-packaged: efficient cooling, renewable offsets, and community engagement.

Worse, tech giants have even started telling reporters that their facilities are “less thirsty” than before, part of a broader industry effort to get ahead of mounting public anger in the US over how much water AI infrastructure consumes.

Also Read: The AI server boom in Southeast Asia: Why data centres are running out of power

That messaging has not yet reached Gelang Patah, a town in Johor, Malaysia, where residents gathered outside a data centre construction site earlier this year holding a straightforward complaint: there was not going to be enough water left for them. It is a small protest that points to a large problem, and Southeast Asia’s AI boosters would rather not dwell on it.

Johor is the test case, and it is already straining

Johor is Southeast Asia’s fastest-growing data centre hub for a reason that has nothing to do with Malaysia’s own digital ambitions. When Singapore froze new data centre approvals between 2019 and 2022 to protect its limited land and water resources, global operators simply moved their plans across the causeway. Johor’s aggregate capacity has since surged towards 5.8 gigawatts, and the state now hosts dozens of operational facilities feeding off a water and power system that was never built for this scale of industrial demand.

The numbers are no longer abstract. A single 100-megawatt facility can consume in the region of 4 million litres of water a day for cooling alone; Johor officials have described hyperscale sites drawing roughly 200 times more water than an ordinary industrial user.

Regulators have responded by rejecting a meaningful share of new applications, raising industrial water tariffs, and telling the largest prospective tenants to wait until at least mid-2027 for guaranteed water and power connections. One estimate puts committed demand from the state’s data centre pipeline at more than 800 million litres a day, against roughly 140 million litres of infrastructure actually capable of delivering it today.

Malaysia’s federal government has all but admitted the model is unsustainable in its current form. Prime Minister Anwar Ibrahim told parliament this year that Malaysia has stopped approving new non-AI-linked data centres altogether, betting that AI-branded projects still qualify for green-lighting even as electricity demand from the sector is projected to climb to nearly a third of the country’s entire power supply by 2035, up from around seven per cent today.

Malaysia is not the exception; it is the preview

Treat Johor as a warning rather than an isolated case, because the same arithmetic is playing out across the region with fewer headlines. Indonesia’s Java-Bali grid, which carries most of the country’s data centre load, was already running near capacity before a single new AI facility came online, and the grid remains roughly 60 to 65 per cent dependent on coal, a fact that sits awkwardly against hyperscalers’ net-zero pledges. Thailand’s data centre power demand reportedly grew fourfold between 2020 and 2024 while generation capacity crept up by less than a tenth of that.

Vietnam, meanwhile, is attracting hyperscale investment on the strength of cheap land and labour even as parts of the country face weekly power cuts during summer peaks.

Also Read: Breaking into the data centre sector: Beyond technical expertise

A recent Bain and Standard Chartered analysis framed this plainly: the binding constraint on Southeast Asia’s AI-driven growth is not capital or ambition, it is the grid itself, and the region’s transmission and distribution networks have not kept pace with the concentrated, high-value demand that data centres represent. Roughly 35 to 45 terawatt-hours of incremental demand is expected across the region’s hubs by 2030 (Singapore, Johor, Bangkok, Greater Jakarta, Manila, and Batam) landing on infrastructure largely designed for a slower, more distributed pattern of growth.

The speculative capacity problem makes this worse

What makes the resource strain harder to justify is that a meaningful share of the demand driving it may not even be real yet. Malaysia’s Energy Commission has found that data centres were drawing less than half of their declared maximum electricity demand as of mid-2025, prompting officials to flag the likelihood of speculative applications — developers reserving grid capacity and water allocations well ahead of actual tenant commitments, effectively queue-jumping scarce resources against uncertain future need.

That is a familiar pattern from past infrastructure bubbles, and it means some of the water and power tension communities are living with today is being generated by capacity that may never be fully utilised.

This is precisely why Johor’s response, however belated, is worth taking seriously as a template rather than dismissing as friction. The state has rejected close to a third of data centre applications over sustainability shortfalls, mandated a shift towards reclaimed wastewater instead of municipal supply for new approvals, and built a dedicated water reuse programme aimed squarely at the industry.

None of that has fully closed the gap between committed capacity and available infrastructure. But it is a materially more honest starting point than the alternative most of the region has defaulted to: approve first, measure the damage later.

Southeast Asia should not import a problem it can still design around

The uncomfortable truth is that Southeast Asia has a genuine opportunity here that most of the world does not. Its AI data centre boom is still young enough that grid interconnection, water accounting standards and siting rules can be built deliberately, rather than retrofitted after the fact the way the United States is now attempting. The ASEAN Power Grid interconnection and philanthropic clean-energy pledges for surrounding communities are steps in the right direction, but they remain medium-term fixes for a strain that is already showing up in tariffs, deferred approvals and community protest today.

Also Read: The US$5 trillion AI data-centre buildout unleashes the paradox that limits its returns

Governments across the region would do well to stop treating data centre investment announcements as unambiguous economic wins and start asking the harder question Johor is now being forced to confront: what does this facility cost the people living next to it, in water, in power, and in a grid that other industries and households also depend on?

The alternative is a region that spent its AI infrastructure boom exporting the same environmental trade-offs Silicon Valley is only now starting to reckon with — except this time, with far less capacity to say no.

The post SEA’s AI boom has a water problem it cannot offset away appeared first on e27.

Artificial intelligence has dramatically changed the way startups are built.

Today, a founder can describe a product idea, open a tool such as Claude or OpenAI, generate hundreds of lines of code, build a prototype and present it as an “AI-powered innovation” within days. What once required a technical team, months of development and significant capital can now be achieved remarkably quickly.

The barrier to building software has never been lower.

But as the ability to build accelerates, another challenge becomes increasingly important: validation.

AI can generate code, analyse information, summarise complex material and produce remarkably convincing answers. But a convincing answer is not necessarily a correct one. And when AI-generated outputs move from a demo environment into industries where mistakes have real consequences, the difference between “working” and “working reliably” becomes critical.

Healthcare is perhaps the clearest example.

For a healthcare AI startup, the question is not simply whether a model can produce an answer. It is whether that answer is clinically reliable, generated from appropriate data, reproducible across relevant populations and settings, understandable to the intended user, and safe enough to inform a clinical decision.

This is where human-in-the-loop verification becomes more than a safety feature. It becomes part of the product itself.

The clinician is not the bottleneck

There is sometimes an assumption that AI automation becomes more valuable as humans are removed from the workflow. In healthcare, that assumption can be dangerous.

AI can process enormous volumes of information far faster than a human. It can identify patterns across patient records, compare information against large knowledge bases and surface potentially relevant findings. But it does not automatically understand the complete clinical context in which those findings will be used.

A clinician does. That is why the most useful healthcare AI may not be the system that attempts to replace clinical judgement, but the one that augments it.

Regulation is increasingly reflecting this distinction. The US Food and Drug Administration’s January 2026 final guidance on Clinical Decision Support Software clarifies the criteria for certain clinical decision-support functions to qualify as non-device software. One important criterion is whether the software enables healthcare professionals to independently review the basis of its recommendations rather than primarily relying on the software’s output.

Singapore’s Health Sciences Authority has similarly refined its framework for Software as a Medical Device (SaMD) and Clinical Decision Support Software (CDSS), as outlined in its update on SaMD risk classification and CDSS qualification guidelines. Its July 2025 revision added, among other changes, a criterion concerning whether CDSS recommendations are based solely on established clinical guidelines when determining whether software qualifies as a non-medical device.

The message for founders is important: automation does not automatically mean removing the professional from the loop.

Depending on its intended purpose, functionality and risk, software may fall within medical-device regulation or qualify for a non-medical-device pathway. Either way, the product needs to be designed around clearly defined accountability.

The clinician interprets the recommendation, considers the patient’s circumstances and decides whether to accept, modify or reject it.

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

“Accurate” is not enough

AI hallucination is often discussed as a technical problem. In healthcare, it is a product and safety problem.

A generative AI system can produce an answer that is fluent, structured and persuasive while being completely wrong. A fabricated reference, incorrect interpretation of a medical record or inappropriate recommendation could have consequences far beyond a poor user experience.

This means healthcare AI cannot be evaluated simply by asking: “How accurate is the model?”

The more useful questions are: Accurate for whom? Under what conditions? Compared with what reference standard? Using which data? And in which real-world population?

Traditional metrics such as sensitivity, specificity, precision, recall and area under the receiver operating characteristic curve (AUC) remain valuable. But a strong metric on a controlled dataset does not automatically translate into reliable performance in clinical practice.

A model can perform exceptionally well in one dataset and behave differently when exposed to another hospital, patient population, imaging device, documentation style or clinical workflow.

This is why validation must extend beyond the model itself.

It includes the quality and representativeness of the data, external validation, clinical workflows, human factors, usability, monitoring and performance after deployment. For regulated software, lifecycle management, verification and validation, change management and post-market considerations are increasingly important parts of the development process. HSA, for example, maintains a lifecycle-oriented framework for software medical devices alongside its SaMD and CDSS classification guidance.

The new startup moat may be trust

For founders, this creates an important strategic shift.

The competitive advantage of an AI startup may no longer be simply how quickly it can build a model.

If thousands of startups can use the same foundation models and increasingly powerful coding tools, the ability to produce a prototype becomes less differentiated.

The harder question becomes: Can you prove that what you built works?

That proof may become one of the strongest forms of competitive advantage.

A startup that combines AI automation with genuine domain expertise, structured validation, transparent outputs and continuous monitoring can build something considerably more defensible than a product that simply places a large language model on top of an existing workflow.

This is particularly relevant for founders entering regulated or high-stakes industries. Domain experts should not be brought in merely to satisfy an advisory requirement after the product has been built. Their expertise should influence the product architecture, validation strategy, workflow design and definition of failure.

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

Human-in-the-loop should therefore not be viewed as a limitation on AI.

It is a mechanism for making AI deployable.

The best systems will know what to automate, when to request human verification and, critically, when not to provide an answer at all.

From “AI versus humans” to “AI plus humans”

The future of AI in healthcare is unlikely to be a simple contest between artificial intelligence and human intelligence.

It is more likely to be a carefully designed partnership.

AI brings scale, speed and the ability to process enormous amounts of information. Humans bring contextual understanding, professional judgement, ethical responsibility and the ability to challenge an output when something does not look right.

The real innovation lies in designing the interface between the two.

As AI lowers the cost and time required to build software, founders will increasingly be judged on something beyond how quickly they can produce a demo.

They will be judged on whether they can demonstrate that their product works, understand where it can fail, and build the mechanisms to detect and manage those failures.

In an age where almost anyone can build software with a prompt, building is becoming easier. Proving is becoming harder.

And for high-stakes AI, that may be where the real startup advantage lies.

—

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 missing layer in AI innovation: Human verification appeared first on e27.