A photo from the soft launch of the report
Southeast Asia’s Generative AI (GenAI) startup ecosystem is vibrant and diverse. Startups use various foundational models to innovate and solve complex problems.
Foundational models are large, pre-trained AI models that can be fine-tuned for specific tasks. Trained on vast datasets, they offer versatility and power for various applications and can understand and generate text, images, and other data forms, serving as a strong base for specialised AI systems, like GPT for text and DALL-E for images.
GenAI Fund, a new investment vehicle in Vietnam, conducted a survey to discover the current trends and opportunities for GenAI Startups in ASEAN, including the foundational models they utilise. At the time of the soft launch, the VC firm analysed over 500 GenAI startups and conducted interviews and surveys with over 180 GenAI startups across the region on 25 different data points.
Also Read: Is generative AI the game-changer for productivity?
Below are the three key takeaways based on the survey’s preliminary findings.
The full ASEAN GenAI Startups report will be released in the coming weeks.
Key Takeaway #1: Vietnam is second to Singapore in GenAI adoption
Based on the startups surveyed up until the soft launch, over 90 per cent are startups at the application layer, with 47 per cent of startups based in Singapore, 32 per cent in Vietnam, and 8 per cent in Indonesia.
While Singapore’s lead is unsurprising for the usual reasons (business/tax structure, funding access, diverse talent pool and network, robust infrastructure and existing startup community, supportive and solid government), Vietnam’s #2 spot is eye-catching.
Anecdotally, it’s worth noting that an increasing number of regional startups (excluding Vietnam) have built/are building their initial MVPs — or established their remote development teams — in Vietnam. The names include Eklipse.gg, a GenAI tool automatically converting lengthy streams into viral clips.
“What is surprising at first glance is Indonesia’s relatively slower uptake in building GenAI startups. When interviewed by us, notable startup founders and VCs attribute this to Indonesia’s penchant for startups that can better leverage its large 280 million population, i.e., primarily B2C startups like GoTo, Traveloka, Akulaku, Kopi Kenangan, etc.,” says GenAI Fund’s Partner, Denning Tan.
However, a number of bright sparks are appearing from Indonesia, such as Meeting.ai, a leading B2B GenAI-powered transcription and summarisation software.
KeyTakeaway #2: The rise of B2B GenAI startups
Prior to GenAI going mainstream (i.e., before the launch of ChatGPT in November 2022), an estimated 80 per cent of ASEAN startups were focused on B2C business models and applications.
Fast-forward to July 2024. Based on the GenAI Fund survey, over 90 per cent of GenAI-native startups are B2B-focused or B2B-led (with some aspects of B2C), which is a complete contrast to non-GenAI startups, generally.
Consistent with B2B SaaS startups, 81 per cent rely on subscription-based pricing models, and 78 per cent have either achieved early recurring revenue, are profitable or have paid pilots and POCs despite 92 per cent being at pre-Series A stage, with an overwhelming majority focused on productivity-based solutions and workflows across several sectors.
Layered on this is that 71 per cent of surveyed startups target a global (mainly US) market from Day 1, while 29 per cent are focused on ASEAN.
This dynamic poses some challenges.
- While enterprises in ASEAN are showing an increased appetite for third-party GenAI/AI-based solutions, the processes to onboard and implement SaaS are still beset by slow processes like Request for Proposals (RFPs), other tender processes, stakeholder challenges, internal resistance, insecurity, excessive POCs etc. This could spell the death knell for fast-moving GenAI startups needing quicker cash flow or traction to prove product market fit.
- For B2B startups targeting a global market and whose product is designed more for larger enterprises, making breakthroughs in a large foreign market without sufficient traction, credibility, and on-the-ground SaaS sales teams/partners is extremely difficult amidst intense competition.
- It does not help that startup funding for ASEAN GenAI startups has been somewhat muted—but that’s a separate discussion.
Recognising these challenges, Big Tech companies like AWS, GCP, and Azure have been initiating programmes to help accelerate go-to-market for GenAI startups.
Also Read: From Amazon to AI: How GenAI Fund fuels innovation in SEA through a unique model
Some startups have shown resilience and are thriving; an example is Singapore-based business productivity GenAI startup Bluesheets, which connects and automates an organisation’s data across the enterprise customer’s existing platforms.
Key Takeaway #3: GenAI Startups have become more sophisticated
How GenAI startups consume tech (in particular, choice of LLMs and cloud) points towards a more sophisticated multi-model and multi-cloud approach.
Starting with LLMs, from GenAi Fund’s survey, 74 per cent of startups currently use Open AI, followed by open-sourced models Llama (Meta) at 39 per cent, Hugging Face at 32 per cent, Claude (Anthropic) at 29 per cent, Gemini (Google) at 17 per cent, Stability AI at 15 per cent, followed by sub-10% usages for other models, including custom/in-house ones.
Practically all are multimodal in approach, displaying sophistication in leveraging open-sourced options for fine-tuning, retrieval-augmented generation (RAG), and other grounding customisations, as well as for simpler reasons like cost optimisations.
“When speaking with foundation model companies, there is a shared consensus that the segment may be commoditising quickly, which also explains the frequent downward adjustment of API prices,” he adds. “Similarly, with cloud consumption, we are seeing an increasing number of GenAI startups adopting a multi-cloud strategy from early stages, as opposed to relying on single cloud architecture for most startups pre-GenAI era.”
Also Read: Experts advocate thoughtful regulation for the rapid rise of Generative AI
As expected, the top 3 cloud providers used by the GenAI startups surveyed are led by AWS at 65 per cent, GCP at 45 per cent and Azure at 43 per cent. One data point here that stood out was the rise of Azure as a cloud of choice for GenAI startups; Azure is traditionally more entrenched with enterprises – aided in part by the popularity of Open AI, and the presumption that a large portion of B2B SaaS GenAI startup’s potential customers are indeed enterprises who use Azure or at least Microsoft 365.
As a result of the perceived commoditisation of both the choice of models and cloud, Big Tech companies are upping the ante with programmes and benefits that specifically target GenAI startups to encourage long-term build and familiarity with their respective products.
Some examples include AWS GenAI Spotlight APJ, Generative AI Accelerate Programme (NUS Enterprise), and Google AI Accelerator Programme (Singapore).
—
Image Credit: 123RF.
The post 3 key takeaways from the soft launch of GenAI Fund’s ASEAN GenAI Startup Report appeared first on e27.