Anthropic is strengthening its infrastructure push with a move that says a lot about where the AI race is heading. Sana Ouji, who until now was a strategic investments and energy alliances executive for data centers at Google, has left the company to join Anthropic’s new energy team. Ouji herself explained on LinkedIn that she will work alongside Ariel Horowitz and Tim Hughes on a global energy strategy aimed at “scaling responsibly and swiftly” an ambitious portfolio of data centers.
This hiring isn’t happening in isolation. According to DatacenterDynamics, Anthropic has been recruiting for months individuals with experience at Google and in the data center industry to build a stronger internal structure. Among the named hires are Winnie Leung, leading data center infrastructure; Liwen Mao, responsible for data center design; Adam Johnson, in electrical engineering; Zach Miller, in operations; plus other technical and security profiles with extensive backgrounds at Google. They are joined by executives like Tim Hughes, from Stack Infrastructure, and Matt Wanner, also from the same company.
The key isn’t just who joins, but for what purpose. Anthropic no longer seems satisfied with merely being a major consumer of external cloud services. It is building internal muscle in energy, capacity, design, execution, and operations—aiming at a much bigger goal: securing power supply, land, capacity delivery, and operational control at a moment when access to computing has become a competitive advantage as crucial as the models themselves. The company’s public job openings reflect this, with roles like Data Center Portfolio Planning & Execution Lead, Senior Data Center Capacity Delivery Manager, Data Center Electrical Engineer, and Data Center Engineer, Resource Efficiency – Compute Supply.
That organizational reinforcement aligns with moves Anthropic announced in recent months. In November 2025, the company revealed a $50 billion investment in compute infrastructure in the US alongside Fluidstack, with data centers in Texas and New York, and more locations planned. At that time, the company stated these facilities would be tailored to their workloads and designed for maximum efficiency. Fluidstack confirmed the deal, noting it would generate hundreds of permanent jobs and thousands during construction.
Adding to this is the dimension of chips and contracted capacity. Reuters reported in October 2025 that Anthropic signed an agreement with Google for over 1 GW of TPU-based compute capacity, expected to be operational in 2026. By April 2026, Anthropic officially announced a new deal with Google and Broadcom for several gigawatts of next-generation TPU capacity starting in 2027. While the company didn’t specify an exact figure, other sources estimate this expansion around 3.5 GW, reflecting the scale of the leap.
In other words, Anthropic is combining three strategic layers simultaneously. The first is the contracted cloud with hyperscalers, maintaining high-volume relationships with Google and others. The second is dedicated infrastructure, like the deployment with Fluidstack. The third is building an internal team capable of negotiating energy, planning portfolios, executing deliveries, and operating capacity—like a large data center company. This mix suggests that the AI battle is no longer just about models, but also about megawatts, supply chains, and specialized talent.
The timing is also notable. The pressure to secure computational resources has become a major sector tension point. OpenAI, in an internal memo leaked to The Verge, accused Anthropic of a “strategic error” for not securing enough compute capacity, which—according to that report—has impacted availability and reliability. While this is a self-interested critique, it highlights how the AI discussion has shifted from labs to the realm of heavy-duty infrastructure.
Interestingly, Dario Amodei has publicly defended a cautious stance regarding unrestrained computing purchases. In February, he warned that if a company misjudges growth by a year or if the market grows five times faster than expected instead of ten, it could go bankrupt by acquiring too much capacity too early. This underscores the delicate balance Anthropic now faces: ensuring enough energy and data centers without overbuilding in a fast-moving, hard-to-model market.
Hence, Sana Ouji’s recruitment carries more significance than it appears at first glance. She is not just a management changeover; she’s a key piece in a broader reorganization: Anthropic is professionalizing its energy and data center infrastructure with profiles from Google, the electric sector, and large data center development firms. The message is clear. To compete at the top tier of AI, it’s no longer enough to have great models—you must secure ground, power, chips, delivery timelines, and teams that can turn massive contracts into real capacity.
Looking at the bigger picture, this move also says something about the current state of the market. For years, tech startups boasted about being lightweight and relying on others’ infrastructure. Generative AI is reversing that logic. Companies like Anthropic increasingly resemble large infrastructure operators: negotiating gigawatts, hiring data center veterans, planning portfolios, and measuring their future in available electrical capacity. This transformation could prove as decisive for the sector as any leap in model performance.
Frequently Asked Questions
Who is Sana Ouji and why is her hiring by Anthropic significant?
Sana Ouji worked over six years at Google and recently held a position related to strategic investments and energy alliances for data centers. Her move to Anthropic strengthens the company’s new energy team amidst rapid infrastructure expansion.
What is Anthropic building in terms of data centers?
In November 2025, Anthropic announced a $50 billion investment in AI infrastructure within the United States along with Fluidstack, which includes data centers in Texas and New York, with additional locations planned.
What is the relationship between Google and Broadcom with Anthropic’s expansion?
Google signed a deal with Anthropic in 2025 for over 1 GW of TPU-based capacity, and in April 2026, Anthropic announced a further agreement with Google and Broadcom to access several gigawatts of next-generation TPU capacity starting in 2027.
Why are energy and data centers now so critical for AI?
Because training and deploying advanced models require enormous amounts of computing, which depends on chips, data centers, electrical capacity, and delivery timelines. Competitive advantage now hinges not only on software but also on the physical infrastructure that enables it.
via: datacenterdynamics