Google has been busy making friends in Singapore. On 7 October, its AI Futures Fund teamed up with Granite Asia to write cheques of up to US$2 million into Asian AI startups. Eight days later, the same fund has walked into Sea Limited’s office carrying something different: not a term sheet, but a key to Google’s model cupboard.
Sea, the Singapore-founded parent of Shopee, Garena and Monee, announced on Thursday that Sea Labs, its technology innovation and development arm, will work directly with Google AI Futures Fund to explore and build new applied AI products.
Teams across Sea will get access to selected frontier models from Google, along with Google experts who can help develop and test Sea’s new product ideas.
Also Read: Granite Asia, Google AI Futures Fund team up to back Asia’s AI founders
Jonathan Silber, co-founder and Director of Google AI Futures Fund, called Sea Labs the fund’s “first partnership with an external AI lab”. It is the most interesting line in the announcement. It is also the one that raises the most questions.
What is actually being traded
The AI Futures Fund sits inside Google Labs and usually backs startups with a bundle of early access to Google DeepMind models, capital, distribution and mentorship. Sea is not a startup. It is a NYSE-listed company that booked US$22.9 billion in revenue in 2025. It does not need Google’s cheque, and the announcement does not mention one.
No investment amount, equity stake, compute commitment, exclusivity terms, timeline or product milestones were disclosed. The currency here is access, and it flows both ways.
Google gets what every model maker craves: a live testbed of hundreds of millions of daily users across Southeast Asia and Latin America, shopping, gaming and borrowing in dozens of languages, on budget phones and patchy networks. Sea gets hands-on time with Google’s most advanced models, plus engineers who can help it decide what is worth building on top of them.
The logic holds. At the API level, frontier models are steadily becoming interchangeable. What separates winners is how fast a model turns into a product people use every day, and few companies in the region ship consumer products at Sea’s scale.
Chapter two, with chapter one still unread
The tie-up builds on the expanded strategic partnership the two signed in February 2026, which covered gaming, agentic commerce and agentic payments. Back then, Shopee and Google said they would explore an AI shopping agent spanning both platforms, while Monee agreed to give feedback on Google’s Agent Payments Protocol (AP2).
Also Read: Sea’s blockbuster 2025 masks a tightening profitability squeeze
Thursday’s announcement offers no update on how those February pilots have fared. Instead, it adds a new layer on top. For a company that wants investors to see it as an AI story, that is a little like announcing a sequel before anyone has watched the first film.
Forrest Li, Sea’s Chairman and CEO, framed the move as part of a bigger shift. “Our ambition to become an AI-native company goes beyond improving what we do today,” he said, adding that some of the most meaningful applications of AI “are still to be discovered”.
The R&D question
That ambition sits a little awkwardly next to Sea’s own numbers. Its research and development expenses fell 7.8 per cent in the fourth quarter of 2025 to US$277.6 million, even as group revenue rose 38.4 per cent. Shopee’s operating margin in that quarter was roughly 2.5 per cent.
Read that way, borrowing a frontier lab’s brainpower is a sensible move for a thin-margin commerce business. It buys frontier capability without frontier capex. The announcement does point to Sea’s own in-house AI stack, including proprietary models and accelerated compute infrastructure, so this is not a company starting from scratch.
The risk is dependency. Every prototype Sea’s teams build on Google’s models makes switching later more expensive. Founders in the Granite Asia programme were explicitly promised full technical choice across their AI stack. Sea’s announcement makes no such statement.
Everyone has a model patron now
Sea is hardly alone in pairing up with a big AI player. Grab, its closest regional rival, has worked with OpenAI since 2024 and last week signed a framework agreement with SoftBank and PETROS to explore an AI infrastructure platform in Sarawak. In Indonesia, GoTo co-developed the Sahabat-AI language model with Indosat Ooredoo Hutchison.
In the region’s e-commerce fight, Lazada draws on parent Alibaba’s Qwen models, and TikTok Shop sits on ByteDance’s in-house AI. Every major Southeast Asian platform now has a foundation-model patron. Sea’s choice of Google fits Google’s own regional playbook of scaling Gemini through schools, telcos and local languages.
The difference is depth. Most of these tie-ups are supplier relationships. Google calling Sea Labs a fellow “lab” suggests it wants something closer to co-development, with Sea’s markets as the proving ground.
What to watch
The announcement promises products tailored to local contexts and needs. In practice, that means multilingual customer service, tools for small sellers, and fraud detection for first-time digital borrowers. These are areas where Southeast Asia’s messiness is a genuine advantage for whoever learns from it first.
Also Read: Vietnam’s shoppers are asking AI first, but Google still gets the last word
But the question raised in February remains unanswered. When an AI agent becomes the storefront, whose interest does it optimise: the buyer’s, the seller’s, or the platform’s take rate? A lab partnership with no disclosed guardrails does not settle it.
For now, “two labs” building together at the frontier makes for a good soundbite. The real test is whether, by the next 12.12 sale, a seller in Surabaya or a gamer in São Paulo notices anything different.
The post Sea and Google open chapter two of their AI pact before chapter one ships appeared first on e27.
Discussion about this post