YC recently updated their request for startups, and it’s helpful to look at this list to see what can be built out in SEA and when SEA builders and investors should consider charting a different path due to their inherent strengths and weaknesses of building in this region.
1. Applying Machine Learning To Robotics
This is an area where SEA founders who have expertise can hire cheaper yet effective engineers to attack complicated engineering problems, giving them more shots on goal for the amount of money raised. The main risk is that the customers are unlikely to be local. Robotics replacing expensive labor is far more viable than robotics replacing cheap labor. Factories in low cost of living countries have less of an incentive to invest in capital that replaces labor, and investing in capital when it is at the bleeding edge of state of the art technology is taking on a lot of risk for minimal upside. On the plus side, there are potential customers in the same time zone in countries such as Korea, Taiwan and Japan.
2. Using Machine Learning To Simulate The Physical World
There is no reason a realm of strong ML engineers couldn’t take this approach in SEA, the main issue is there are fewer clusters of ML engineering teams in this region. But there are sizable groups in Singapore working on ML and there are pockets of strong AI talent in places like Hanoi, so it’s plausible that someone starts something interesting in this space.
Building defense oriented companies is trickier in SEA, as these countries will be regulating defense tech far more closely than they are tracked in the US. While the US purchasing system has various inefficiencies, they are a big buyer and they have the best technology. And when the US is not a likely customer because they are buying domestically, it makes building for this space very tough. But if Baykar can make a billion dollars in revenue building drones based out of Turkey, there is opportunity to do this in other parts of the world. The region and the world needs more capabilities to help counter an increasingly aggressive CCP. But US investors should realize that the political and regulatory risks of playing in this space are far larger when they venture outside The Gundo and the rest of the US.
4. Bring Manufacturing Back To America
There is no way an SEA startup will truly be focused on manufacturing in the US. But there is a broader trend of China + 1, in which companies are figuring out how to move at least some of their production outside of China. It’s far more viable for them to relocate that production to parts of SEA than it is to the US. The question is what type of business or technological solution to the China + 1 problem will generate venture scale returns in this space. The FDI is going to be coming into Vietnam, Indonesia and other markets as businesses realize they cannot rely only on China. The question is whether technology companies can be uniquely helpful enough in the process to capture a significant portion of the value they are enabling.
The fastest wins in the manufacturing space will be from experienced teams that bring in advanced manufacturing, as local governments are more likely to subsidize these ventures, and non-dilutive funding always helps.
An orthogonal but important point is that many of the companies that have been manufacturing things in the world of atoms still use China for their production. Switching outside of China is a luxury for profit making businesses that want to protect their downside. Unless it is the point of the company to relocate manufacturing it is not necessary to incorporate this idea.
There have been a couple space companies built out of Singapore. It’s an exciting space, but so far the end user economics is either tied to telecommunications, government and defense related work, or capturing money being spent by eccentric billionaires who want to mine and colonize the solar system. The more viable companies will be building for the first two types of customers. When building or investing in space-tech, know your market to make sure that your market size is adequate.
6. Climate Tech
Climate tech in SEA is an interesting beast. Many of the opportunities are coming from regulations from outside of SEA. The potential generation of carbon credits from Indonesia for European companies is enormous, and a well connected local team can start a company out of the gate facilitating an amount of carbon credits rivaling that of much better funded US startups. The question is where things go from there and if players in this region can develop scalable business models or compete on the technology side. As a financially motivated investor, it feels very dangerous to play in spaces occupied by many people who are deploying money for non-economic reasons. There are local subsidies and incentives for EVs, but it’s not obvious that there is product market fit among the end users. Those who are not careful risk ending up like Ford, losing $4.7 billion dollars in 2023 trying to make EVs work while not coming close to capturing the market.
7. Commercial Open Source Companies
This is an area where SEA’s lower cost and high quality engineering talent can really shine. There are a lot of really strong engineers and founders in the region. The challenge comes from the sales and distribution side. India has a BPO and SaaS sales culture in place, but outside of the Philippines, where engineering talent is not as dense, it’s hard to find people willing to make themselves available on Western time zones. Founders who are tackling this problem from SEA need to have feet in both worlds to have the best shot at making this type of company work. And any company that doesn’t have a smart distribution strategy is going to languish unless they really did build the best tools in their space and are lucky enough to be getting organic traction.
AR/VR in the context of building for Apple Vision Pro, Meta Quest 3 or other devices is still early. SEA founders can use their engineering cost advantage to build apps and tools for these environments. It’s an exciting market in part because we don’t yet know what the product market fit looks like outside of a few niche areas. It’s also a distant market, because there really isn’t any major PMF that we can point to. Apple and Meta are going to keep investing, so it’s likely that we will see a few more small wins in this space even before we figure out what the future looks like. But building to be bought out is always a risky move (unless you are in a space like biotech where the dynamics are more broadly understood - but even in this space the upside is capped relative to the rest of the tech).
9. New Enterprise Resource Planning Software
This is an area that might be approached in SEA from a fast follower and fragmented world market perspective. The risky part of building this primarily in SEA for the SEA market is that the multinationals, which make up large portions of potential customers, tend to prefer using the same tech stack throughout their company. Still, the market is large and deep enough that we should expect a few winners. One encouraging trend in Indonesia is that startups have moved beyond the stage where they give away their SaaS solutions to companies and try to make up for it by charging for ancillary services. Building for the world market encounters challenges related to selling to the world from SEA, so any founder building for this market again needs to be sure they have the context of the problems they are solving and the ability to spend time to form the relationships needed to make sales happen.
10. Developer Tools Inspired By Existing Internal Tools
This is an interesting concept that doesn’t apply as directly in SEA, where internal corporate tools are not held in as high regard as those emerging from FAANG (Or MAMAA or the Magnificent Seven or whatever we are using as an acronym for the elite tech companies of the United States). The approach that might work better in SEA is solving common business operational problems inspired by internal issues at top local companies.
11. Explainable A.I.
This seems like a feature of a successful AI based company more than the end in itself, but making sure processes are explained is a useful concept for people doing important work with AI models and other algorithms. Generating multiple similar results with different types of approaches can reduce errors and create confidence for people who will need to trust their outputs. While LLMs can do a lot of the heavy lifting, it’s interesting to think about areas where alternative processes can act as a double check.
12. L.L.Ms For Manual Back Office Processes In Legacy Enterprises
Service Now has been seeing a lot of growth in the US as they transition their traditional back office offerings to AI enabled processes. There is a lot more that can be done in this area. The SEA’s fragmented market presents challenges and opportunities, so companies who figure out this approach for the US will be reluctant to expand to SEA quickly. This is an area where a fast follower approach that gloms onto a working model, modifies and applies it to SEA before the companies decide to come out to this region could be very successful.
13. A.I. To Build Enterprise Software
This is a tough idea, as most enterprise software needs to be extremely reliable and current LLM approaches still have lots of hallucinations. But outside of execution, the main challenge for founders building in this space in SEA is either needing to sell it to local customers, or making sure they are global enough to sell to Western customers. One fun trick with selling globally from SEA, for those who are not yet hiring local sales reps, is to make sure that your availability looks similar to that of your localized competitors. Offer 9 to 5 hours in their timezone, even if that means taking a call at 3am.
The initial iterations of this style of custom software will still be relatively formulaic, but as programmers get faster at creating working custom software the dynamic of enterprise sales really does change.
Stablecoins are an interesting regulatory arbitrage - they can be money market funds that avoid actually registering as funds. And with the broader crypto ecosystem still not expecting significant yields after various firms lost all of their funds chasing yield, these offerings should have a high net interest margin and be quite profitable. There might also be good reason to expect more of these companies in SEA, as it makes sense to build these products away from the jurisdictions of the SEC and other US regulatory bodies. US regulators might be losing in court, but they are still quite antagonistic to crypto.
MRIs have an interesting potential to detect abnormalities. And there are many other approaches, finding markers in blood or in urine that AI tech can detect cancer. This is an area where serving the US market is far more lucrative than trying to sell in SEA. There may be ways to do testing more efficiently in Thailand’s medical ecosystem, but outside of something interesting coming out of Singapore’s university system we don’t expect there to be many opportunities in this space or related spaces in the region.
Still, founders should keep in mind that there are interesting things that might be done with Thailand’s well developed medical ecosystem.
16. Foundation Models For Biological Systems
Foundational models are very capital intensive. Capital intensive businesses in the healthcare space are generally not the right fit for SEA based startups. Capital is harder to access in the region, so after blowing through some capital the next rounds will remain difficult if they don’t produce amazing results. And the human body works mostly the same anywhere, there are no obvious cost or business advantages to doing this out here. It is only once copyright or data access concerns start slowing down LLM training in the US then it would make sense to explore creating models in jurisdictions where access to the relevant data is easier.
Some interesting specialized AI companies have been formed in Australia and Singapore using more traditional medical imaging and other data. Eventually, the biggest models will benefit from massive amounts of data on patients, interventions and outcomes in ways that it would make sense to focus on parts of the world where the legal system facilitates this data usage. However, external investors should not imagine SEA to be a mythical land of open data and no medical privacy regulations. Activists trained in the West have been pushing for the types of data protections that are hamstringing the emergence of functional AI doctors in the West.
It will make sense for startups in the healthcare space to piggyback on other company’s foundational models and form operating companies to capture the opportunities in this market. Even those opportunities will be complicated by the local regulatory landscape, but once AIs are helping us understand the human body at a level of biological researchers and doctors there are many amazing opportunities.
17. The Managed Service Organization Model For Healthcare
Managed service models might be an interesting opportunity in the US, but while the US is finding alternatives to classic PE plays, these plays may make more sense in SEA. Doing so in a way that is operationally sound, complying with regulations and keeping customers happy while making a profit, is the main challenge. Startups have also found some success building out plays that can be justified on PE type metrics - SEA has always been a region where PE and VC investment tend to overlap. In Vietnam, Nhi Dong 315 figured out a healthcare clinic model for mothers and babies and had a model in Vietnam that attracted significant investment as the investors figured they could copy and paste the clinics around the country. Jio Health was able to do something similar for primary care, chronic diseases and pediatrics.
One aspect of MSOs is that value is created in part by sourcing patients for the organization. There are many founders interested in building on the sourcing side, and the most obvious opportunity is to channel patients to medical tourist destinations in Thailand and elsewhere in SEA. The costs are low and the quality is relatively high, so matching patients with providers creates an obvious win-win-win, but unless the founders find a new model that is large and defensible with patients around the world, it is unlikely that these opportunities will scale to the size that can justify VC investment. In order to reach that type of scale, founders need to help solve the problem of getting people treatment inside the country where they live.
18. Eliminating Middlemen In Healthcare
Many aspects of the healthcare systems in SEA countries are not nearly as cost inefficient as they are in the US. Singapore’s regulatory regime causes drug prices to be much higher than they are in surrounding countries, but it’s the exception.
What is missing is the next step in telemedicine - AI enabled healthcare. Unfortunately, there is no country in SEA where doctors do not have some significant amount of political power. They will fight attempts to replace them. But there is room for AI enabled tools that supplement doctors, take notes automatically, generate potential diagnosis, guide appropriate followup, etc. Hallucinations, inaccuracy and reliable access to high quality training data are among the many challenges in the space, but we are closer than ever to a world where high quality medical advice is within reach of everyone.
Companies that make it easier to customize code and connect different products will benefit from being based in a geography where engineering talent is cheap relative to their strength. As mentioned above, the sales and distribution issues are the challenges that the SEA founders who want to build in this space will need to think through.
20. Small Fine Tuned Models As An Alternative To Giant Generic Ones
The open source LLM ecosystem is rapidly catching up with the state of the art closed source companies. And cheap human labor may be just what is needed to curate data for these models. So this is definitely something that SEA’s AI developer can look into building here. But they also need to be realistic about the funding disadvantage that they face in this market. Absent an enthusiastic sovereign backer or access to somewhere with extremely cheap power to reduce training costs (didn’t the bitcoin miners find most of those?), it will be a more difficult area to build out of SEA. Some biotech companies are formed in Singapore and relocate to a US biotech hub once they get enough funding. We might see something similar for promising AI companies.
For those who want to build AI from SEA, there are plenty of interesting developer tools that can need to be built, and if they stay away from training relatively large foundational models there are many tools that AI developers and companies will need.
Concluding thoughts:
Fisking is usually done aggressively. Not aggressively as in replacing the k with a t, but it is usually done as a critique. In this case, it was done to give those building an investing in SEA, or those who are interested in the region, a perspective on what might be easier or harder to build out here.
Reading through YC’s request for startups emphasizes just how many really cool things the world needs right now. When founders in SEA are deciding what they are building and who they are building it for, they need to be honest with themselves about their comparative advantages and disadvantages. With the recent AI boom, a lot of innovation became centralized in Silicon Valley. But the tools and methods that are powering this trend are available for technologists everywhere. While founders in SEA face tougher fundraising environments and don’t have as much access to talent from top global companies unless they are from those companies themselves, they do have lower costs, faster growing local markets, different legal regimes (a blessing and a curse), and the opportunity to fast follow emerging trends in the US that won’t hit local markets for years. As long as they are thoughtful about the advantages and disadvantages of building in SEA, there are many ways they can create huge wins.