Google Cloud AI Strategy: Why Thomas Kurian Believes Custom Chips Will Beat Amazon and Microsoft
The Google Cloud AI strategy is finally hitting its stride, according to CEO Thomas Kurian, who argues that the company’s pair of newly unveiled chips combined with rapid breakthroughs at its DeepMind research lab will help it close the substantial gap with cloud computing leaders Microsoft and Amazon. After years of trailing competitors and watching rivals dominate the artificial intelligence conversation, Google believes its in-house approach is finally paying serious dividends.
The Full-Stack Advantage
Kurian made his case clearly during a recent interview, emphasizing that Google’s full-stack AI strategy sets it apart from competitors. The approach involves building chips, data centers, foundation models, and customer-facing products entirely in-house rather than depending on outside vendors for critical components.
He directly contrasted this with rivals, arguing that Google isn’t just a hyperscaler reselling someone else’s technology. The company owns its intellectual property, its AI models, and its semiconductor designs, which creates fundamental economic advantages.
The math behind this argument is compelling. Kurian explained that for every dollar of revenue Google Cloud generates, the company isn’t sending 80% of it to outside model or chip providers. That retained capital can be reinvested into further development, creating a virtuous cycle that competitors using third-party AI infrastructure simply can’t match.
A Growing Contender
Eight years after joining Alphabet from Oracle, Kurian has steadily grown Google’s cloud market share from 7% to 14%, doubling the company’s footprint in a fiercely competitive industry. That trajectory has positioned him as a strong candidate to potentially lead Google itself someday.
However, Google Cloud remains firmly in third place behind Amazon Web Services and Microsoft Azure in the massive $418 billion cloud computing market. The company has also faced criticism for allowing chatbots and coding assistants from OpenAI and Anthropic to leapfrog its own AI offerings in consumer mindshare and enterprise adoption.
Recent Numbers Tell a Better Story
Despite the challenges, Google Cloud is suddenly growing faster than its larger rivals. The division reported a remarkable 48% jump in revenue during the final quarter of 2025 and remains on track to generate more than $70 billion this year, up significantly from $43 billion in 2024.
That kind of acceleration suggests the AI investments are starting to translate into actual customer adoption rather than just impressive technology demos.
Custom Silicon Versus the Competition
Google strongly believes its TPUs (Tensor Processing Units) and Gemini AI models substantially outperform what Amazon offers with its Trainium chips and Nova AI system, as well as what Microsoft provides through its Maia processors and MAI models. This positioning matters because it makes Google less dependent on partnerships with companies like Anthropic and OpenAI, while also reducing reliance on Nvidia’s expensive GPU chips.
Kurian credited Google’s 12-year investment in DeepMind for enabling continuous improvements to proprietary chips and the ability to deliver both consumer and enterprise AI products at lower costs with better profit margins than competitors achieve.
The Las Vegas Reveal
This week in Las Vegas, Google unveiled two new chips at a major industry event. The releases represent the eighth generation of TPUs, with the chips specialized for different functions. One focuses specifically on training AI models, while the other features more memory designed for running AI systems faster during what’s called inference, the actual real-world use of trained models.
According to Kurian, building genuinely impressive chips requires a large in-house research lab, and he argued that other players in the space aren’t building their own quality models. He suggested that only Nvidia currently rivals Google’s combination of AI hardware and tightly integrated chip software.
Tension With Nvidia
Google’s emergence as a serious competitor to Nvidia has strained relations between the companies, even though Alphabet remains one of Nvidia’s largest GPU customers. A recent report from Epoch AI estimated that Google controls roughly a quarter of global AI computing power, with approximately 3.8 million TPUs and 1.3 million GPUs in operation. Microsoft holds second place with 3.2 million Nvidia GPUs.
The friction has spilled into public discourse. In a recent podcast appearance, Nvidia CEO Jensen Huang criticized Google for not submitting its AI chips to independent testing and questioned the performance and efficiency claims being made. Huang even suggested that 100% of TPU demand comes from Anthropic, implying there would be no meaningful TPU growth without that single customer.
Kurian’s Strong Pushback
Kurian fired back convincingly, noting that nine of the top 10 AI labs currently use TPUs, including Thinking Machines, the company founded by former OpenAI executive Mira Murati. OpenAI itself can’t use TPUs due to an exclusivity arrangement with Microsoft, but its absence isn’t because Google’s chips lack capability.
He emphasized the simple market reality that customers have choices. If Google’s offerings weren’t competitive on performance, pricing, and quality, sophisticated AI labs simply wouldn’t choose them.
The Anthropic Deal Changes Everything
Friday brought a major development that strengthens Google’s position significantly. Anthropic struck a new agreement to purchase substantially more of Google’s chips, while Google committed to investing up to $40 billion in the AI startup and providing 5 gigawatts of computing capacity over five years. The total value of the deal exceeds $200 billion, representing one of the largest commercial AI partnerships ever announced.
This kind of arrangement essentially locks in massive demand for Google’s chips while giving Anthropic the computing resources it needs to compete with OpenAI’s similarly massive infrastructure deals.
Massive Spending Continues
Google is investing heavily in its AI ambitions, with capital expenditures forecast to reach $185 billion this year. Kurian argues these enormous sums are justified by genuine customer demand and the strong revenues already flowing in. The numbers represent some of the largest infrastructure investments in corporate history, reflecting just how seriously Google takes its AI competition.
A Different Take on AI Startup Economics
Kurian offered an interesting perspective on the financial challenges facing OpenAI and Anthropic, suggesting both companies face genuinely difficult paths forward that could create problems for Big Tech firms depending on them. The two AI startups are losing tens of billions of dollars annually while racing to secure the computing power needed to train and operate their increasingly sophisticated models.
He warned that AI providers depend on private capital markets, which he believes are approaching saturation. Going public requires eventual profitability, while staying private indefinitely means perpetually raising venture capital, neither of which represents a sustainable long-term path.
This year alone, OpenAI and Anthropic raised more than $150 billion combined in two of the largest private fundraisings in history as they prepare for eventual IPOs. Dozens of additional startups have completed multibillion-dollar funding rounds, creating an unprecedented capital flood into the AI sector.
Predicting a Shakeout
Kurian predicted that the next year or two will bring meaningful market shakeout, with the survival of particular providers depending largely on their underlying economics. This represents a notable warning from a major industry executive about the sustainability of current AI startup spending patterns.
His perspective carries weight given Google’s unique position as both a major customer and competitor in the AI space, with deep visibility into actual demand patterns and computing requirements.
What This Means for Enterprises
For business customers evaluating cloud platforms, Google’s emerging position offers some interesting considerations:
- Custom chip ownership potentially translates to better pricing for enterprise AI workloads
- Integration between Gemini models and TPU hardware may deliver performance advantages
- Reduced dependency on Nvidia GPUs could mean more reliable computing access during shortages
- Google’s massive infrastructure investment suggests long-term commitment to the cloud business
Final Thoughts on Google Cloud AI Strategy
The Google Cloud AI strategy represents one of the more ambitious bets in modern technology. By controlling everything from chip design through end-user products, Google is wagering that vertical integration creates sustainable competitive advantages that companies depending on third parties simply can’t match.
Whether this strategy actually translates into closing the gap with AWS and Azure remains to be seen, but the recent revenue growth, the massive Anthropic deal, and the new chip generations all suggest meaningful momentum. Kurian and his team have positioned Google Cloud for what could be its most consequential year yet, with the AI infrastructure boom potentially rewarding the company’s long-term investments in ways that finally pay off.
For an industry watching closely whether Google can convert its AI research advantages into commercial cloud success, the next year will provide critical answers about whether owning the entire technology stack delivers the strategic benefits Kurian believes it does.
