She wrote on LinkedIn that she was “incredibly proud” of the work her teams did in the Philippines — the startups that grew into unicorns in Singapore, Malaysia, and here. Then she packed up and went to work for the competition. Meet Sandhya Devanathan, OpenAI’s newly named vice president for Southeast Asia and Australia. And if you still think the AI race is about who trains the fanciest model, she is the reason that story is ending.
>This hire looks like a small personnel item at first glance. Read it side by side with what OpenAI has been doing across Asia-Pacific this year, and it starts looking like the real front line of the AI industry.
What actually happened
Devanathan spent more than a decade at Meta, joining in 2016 when, as she put it, the company had “just about 300 employees in Singapore.” She rose through the Asia-Pacific gaming business, helped expand Meta’s presence in Singapore and Vietnam, and eventually became its vice president and head for India and Southeast Asia — one of the most fought-over stretches of the global internet.
Now she’s crossing to the other side. OpenAI confirmed the hire this week: Devanathan will be vice president for Southeast Asia and Australia, based in Singapore, reporting to Kiran Mani, OpenAI’s managing director for Asia Pacific. It’s a newly-created position. Her remit covers consumer growth, enterprise adoption and partnerships, regulatory engagement, and operations across the region — the whole business, not one slice of it.
She called the departure “bittersweet” and said she’d take a short break before her “next exciting chapter.” A decade is a long time in tech, and she carries more than 25 years of experience across technology and financial services. This wasn’t a paycheck move. It was a relocation of trust.
Watch the pattern, not the headline
One executive changing companies is a Tuesday. The pattern is the story. Just days before Devanathan’s move, OpenAI named Prabhjeet Singh — Uber’s former India head, another decade-plus operator — as its managing director for India, a market Sam Altman has repeatedly called central to OpenAI’s future. And this isn’t a sudden scramble; it’s been quietly building. Over the past two years, OpenAI opened offices in Singapore, Tokyo, Seoul, Sydney, and Delhi, planting a flag in every major Asia-Pacific market before most people noticed.
This is chess, not track and field. The pieces OpenAI is moving aren’t models — they’re people who already know the board. I wrote recently about Jeff Dean leaving Google to automate science; that was the research-talent version of this story, brilliant minds heading off to build. This is the operational version — executives leaving to run a whole region for someone else. Same signal, different front.
The moat moved from models to operations
For two years we’ve been trained to care about benchmarks — token prices, context windows, glossy leaderboards. I’ve written my share of that myself, from Anthropic’s $45 billion compute deal to the debt-fueled buildout keeping the chips flowing. Here’s the uncomfortable truth underneath all of it: a frontier model is a commodity that keeps getting cheaper, while the relationships required to sell AI into a country are not.
Governments don’t buy a benchmark. Enterprises don’t sign a contract with an API key alone. They buy from someone they trust, in a language and a regulatory context they understand. That trust is the asset Devanathan walks out of Meta carrying. She knows which regulators matter in Bangkok, Manila, and Jakarta. She has watched government citizen services roll out on WhatsApp across India. She has seen Southeast Asian startups scale past a billion dollars. OpenAI is paying for the map, not the model.
The other side of the board
There’s a second reason OpenAI is suddenly staffing up across Asia-Pacific, and it’s the one I think gets missed in the American coverage. The region isn’t just a market to be sold to. It’s the middle of the hottest front in the US-vs-China model war. Chinese labs have been shipping capable open-weight models that run cheap and are willing to adapt to local languages and pricing in ways the frontier labs historically weren’t. I walked through how to run one of the big ones — Tencent’s open-source Hy4 — in a recent tutorial, and it’s telling that a 770-billion-parameter model is now something a determined local developer can stand up on their own kit.
Hold that against OpenAI’s strategy and this hire makes even more sense. You cannot out-benchmark the open-weight ecosystem forever; the gap keeps shrinking. What you can do is be the trusted, regulated, enterprise-ready layer on top — the one a government or a bank actually signs. That job requires people who know the account managers, the regulators, and the cultural norms of each country. That is precisely the talent Devanathan represents. The move isn’t just a people play. It’s a hedge against the cheap open models flooding the market.
Why leave a decade-long run now?
The timing deserves a closer look, because Meta is under real pressure in one of its biggest markets right now. Earlier this month, Instagram mistakenly restricted a post from Prime Minister Narendra Modi, and Meta apologized after Indian authorities summoned senior executives including its chief global affairs officer. On top of that, New Delhi demanded answers after a BBC report found advertisements on the platforms allegedly offering access to child sexual abuse material — Meta says it removed 160,000 accounts in India over six months on signals of child-exploitative activity.
I’m not going to psychoanalyze a public resignation from across the ocean. But it’s worth noting that the people getting hired away are exactly the ones who spent a decade as the public defenders of a business under siege. When the company a region’s operators spent years building starts summoning your boss to answer for what the algorithm surfaced, the guy across the trench offering you a clean, newly-created regional title looks more attractive. OpenAI is recruiting at the exact moment its biggest rival is distracted.
Felix’s take: the Philippines is on the board
This is the part that lands close to home. In her goodbye note, Devanathan called out the Philippines by name — the startups that reached unicorn status in Singapore, Malaysia, and the Philippines. As an ICT division manager here, I’ve watched this gravitational pull from the other side of the fence for years.
When a company like OpenAI decides Southeast Asia is worth a senior regional operator, it stops being theoretical. It means OpenAI — and every competitor reacting to it — will build local teams, hire Filipino engineers and partnerships people, and fight for commercial and institutional trust in our enterprises and government. That’s a genuine opportunity: a rising tide of regional AI investment lifts the market for Filipino talent and the urgency around the technologies I manage.
It’s also a warning. Every Filipino operator good enough to be a regional lead at a global AI company will eventually face the same pull Devanathan did. And our own institutions — public and private — are often slow to make staying compelling. I’ve watched good people leave my own field for opportunities abroad. That gravity doesn’t only apply to research scientists in California; it works on regional operators here too.
I want to be honest about the ambivalence here, because it’s not a clean win for us. More regional AI investment means new tools, new training, new chances for Filipino devs to work on hard problems at global scale. But it also means a growing share of how our country adopts AI will be shaped by decisions made in San Francisco or Beijing, optimized for their home markets first. There’s a real difference between building AI capacity here and renting it from whoever wins the regional turf war. As someone whose job is to help steer technology decisions for a public institution, I’d rather we be players on that board — making informed choices about which layer we trust — than just squares waiting for the next move.
What this actually means for the industry
Step back and the picture is clear. The AI industry has moved through phases: first train the best model, then build the best infrastructure, and now win the best regions. You can see the same land-grab everywhere else in the sector — why open-weight AI companies have become the Valley’s hottest acquisition targets, and why Google is burying search results under AI summaries. Everybody is racing to own a surface — a distribution channel, a market, a stack — before the consolidation begins.
From a Filipino vantage point, the practical read is that AI’s local face is about to change fast. Language, regulation, and trust will decide who actually gets to serve the Philippines, not who posts the biggest benchmark. And the people who win those seats will be the ones who’ve spent a decade inside a market — exactly the kind of operator OpenAI just signed.
Devanathan heads off for a short break before her next chapter. The interesting part isn’t where she lands. It’s that the industry has reached a stage where the most valuable move a frontier AI company can make is hiring the person who already runs a region. That tells you where the real chessboard is. The countries aren’t just spectators anymore — they’re squares on the board. And the next move by an AI titan could land right on the Philippine square.
Sources: TechCrunch and The Hindu.