DeepMind Leadership Change: Why Google Cloud Could Be the Biggest Winner?

TL;DR
· SemiAnalysis believes that following the leadership change at DeepMind, Google Cloud may become the biggest financial beneficiary in the short term.
· Alphabet's Google Cloud revenue grew 82% year-on-year in Q2 to $24.768 billion, with cloud business backlog reaching $514 billion.
· The report estimates that over $150 billion in TPU system orders could push Google Cloud's revenue growth rate to around 150% by 2027.
· This forecast depends on order delivery, customer financing, and revenue recognition pace, and the valuation of system sales may also be lower than traditional cloud services.

On August 7, SemiAnalysis released a report stating that after the leadership change at Google DeepMind, Google Cloud could become the largest financial beneficiary in the short term.

The report links personnel changes to the allocation of computing power within Google. TPUs are a scarce resource that Google relies on to develop cutting-edge models and expand its cloud business. More computing power allocated to Gemini training will help Google catch up with leading models; more TPUs directed to external customers can be converted into Google Cloud revenue, backlog, and profits more quickly.

Based on this, SemiAnalysis provides a set of aggressive forecasts: TPU system sales could push Google Cloud's revenue growth rate to around 150% by 2027, significantly higher than the 64% consensus expectation listed in the report, contributing about $3 to Alphabet's earnings per share that year.

DeepMind Leadership Change: Market Begins to Reassess Computing Power Allocation

According to Axios, Demis Hassabis has stepped down as CEO of Google DeepMind to become the chairman of DeepMind and Alphabet's chief scientist; former DeepMind CTO Koray Kavukcuoglu has taken over daily management and reports to Alphabet CEO Sundar Pichai.

At the same time, Jeff Dean, Sanjay Ghemawat, Oriol Vinyals, and Quoc Le have left Google to co-found an AI research organization called Discovery Loop. Google will participate as a founding investor and establish a cloud services partnership with them.

This round of adjustments has triggered a market reassessment of Google's computing power allocation strategy. The same batch of TPUs must support both Gemini training and inference, can also be provided to external AI labs and enterprise customers through Google Cloud, and can be sold in full sets to special purpose entities operating AI data centers.

Different allocation methods correspond to different returns. The computing power invested in Gemini relates to Google's competitiveness in the field of cutting-edge models; TPUs directed to commercial customers can enter Google Cloud's revenue and backlog more quickly.

SemiAnalysis judges that Google Cloud head Thomas Kurian's influence in the internal computing power competition is rising. As more TPUs and supporting infrastructure flow to external customers, Google Cloud may gain greater revenue elasticity.

DeepMind's Share of Google AI Computing Power Declines

However, Alphabet has not attributed this organizational adjustment to obstacles in Gemini's development. The company stated during its Q2 earnings call that Gemini 4 is undergoing its most ambitious pre-training to date, and AI services are still supply-constrained. Pichai also emphasized that the primary task of TPU allocation remains to ensure that Google has the computing power needed to participate in the competition for cutting-edge AGI.

Currently, Google has not announced a reduction in the priority of computing power for Gemini. The statement that "Gemini's priority has declined" is SemiAnalysis's inference based on personnel changes and resource flows, rather than a confirmed strategic adjustment by Google.

Cloud Revenue Soars 82%, TPU System Sales Begin to Rewrite Growth Structure

Alphabet's Q2 earnings report shows that Google Cloud revenue grew 82% year-on-year to $24.768 billion, with cloud business backlog increasing to $514 billion. The processing volume of Google's proprietary model API reached approximately 22 billion tokens per minute, up from 16 billion in the previous quarter.

The rapid growth of Google Cloud does not come solely from traditional cloud services. Alphabet has begun delivering complete TPU systems to customer data centers and confirmed related revenue for the first time in Q2. The company stated that TPU system sales contracts have been included in the cloud business backlog, with most of the revenue from existing contracts expected to be recognized in 2027.

This means that the revenue structure of Google Cloud is changing. In the past, cloud revenue mainly came from ongoing services such as computing, storage, databases, and AI platforms; now, complete TPU system sales are beginning to become a new source of growth.

Direct Sales of TPUs to Anthropic

SemiAnalysis estimates that the revenue recognized from TPU system sales is approximately $35 billion per GW, with related revenue recognized in Q2 2026 amounting to about $1.2 billion. According to their calculations, if this revenue is excluded, the year-on-year growth rate of Google Cloud's core business remains just above 70%; including TPU system sales, the overall growth rate rises to 82%.

The economic attributes of the two types of revenue are not the same. Traditional cloud business mainly depends on customer renewals, computing power usage, and platform stickiness; TPU system sales are closer to large-scale AI infrastructure projects, where once concentrated delivery occurs, it can significantly elevate quarterly or annual revenue growth.

$150 Billion in Orders Awaiting Confirmation, 2027 Growth Rate Could Reach 150%

The most striking prediction from SemiAnalysis comes from the backlog of TPU system orders.

The report estimates that Google Cloud's TPU system-related backlog orders have now exceeded $150 billion. If these orders are delivered as planned and recognized in 2027, Google Cloud's revenue growth that year could reach around 150%, while the consensus expectation listed in the report is about 64%.

Google Cloud Revenue Growth Rate Forecast

The $514 billion cloud business backlog disclosed by Alphabet includes TPU system sales but does not separately disclose the scale of TPU orders. Therefore, the figure of over $150 billion comes from SemiAnalysis's model calculations.

According to the report's revenue recognition criteria of approximately $35 billion/GW, a single GW-level project is enough to significantly change Google Cloud's growth curve. SemiAnalysis also predicts that as AI labs continue to procure large-scale computing power, more GW-level transactions may occur in the coming quarters, adding over $250 billion in TPU orders to Google Cloud's remaining performance obligations.

Profit margins may also receive support. SemiAnalysis expects that the EBIT margin from TPU system sales is slightly lower than the core cloud business's level of just over 30%, but Google Cloud's overall EBIT margin may still maintain between 35% and 39%. With high revenue growth rates and relatively high profit margins working together, this related business could contribute about $3 EPS to Alphabet in 2027.

Thus, SemiAnalysis connects a personnel adjustment in the AI department into a complete financial chain: changes in the Gemini organization affect computing power allocation, more TPUs flow to external customers, system sales expand cloud business backlog, and ultimately enter Google Cloud revenue and Alphabet profits.

Even if Gemini Falls Behind, Google Can Still Profit from the AI Race

SemiAnalysis holds a notably pessimistic view of Gemini's long-term competitiveness, but for investors, the more easily verifiable question through financial reports is: who ultimately uses and pays for the expensive TPU clusters, and whether these orders can be converted into Google Cloud revenue.

The report points out that the token growth rate of Gemini's first-party API has already slowed. In Q1 2026, its processing volume increased from 10 billion to 16 billion tokens per minute, a quarter-on-quarter growth of 60%; in Q2, it further rose to 22 billion tokens, but the growth rate dropped to about 38%.

This does not mean that Gemini's usage has declined, but rather that its growth rate has slowed. Google officials still state that demand for model APIs is strong, and the company faces supply constraints in computing power, with pre-training for Gemini 4 continuing to advance.

Gemini First-Party API Revenue and Growth Rate

Meanwhile, Google's enterprise AI platform not only supports Gemini but also provides customers with third-party models like Claude. Even if Google's self-developed models are no longer always leading, Google Cloud still has the opportunity to capture the capital expenditures of the entire AI industry through the sale of computing power, platform services, and TPU systems.

Discovery Loop may also enter this cycle. Google is not only a founding investor in this organization but will also establish cloud services partnerships with it. SemiAnalysis further speculates that researchers leaving Google may obtain financing from external investors and related capital from Google, and then use some of the funds to procure computing resources on Google Cloud.

This brings an counterintuitive business result: even if Google faces pressure in the competition for cutting-edge models, it may continue to earn revenue from external AI investments through its cloud infrastructure.

Investor Metrics: Behind the 150% Growth Rate, the Value of TPU Orders Remains to Be Tested

TPU system sales can quickly boost Google Cloud revenue, but how highly this revenue can be valued still depends on how the market assesses its sustainability.

Traditional cloud business relies on customer renewals and usage growth, with relatively stable revenue, and platform migration costs can enhance customer stickiness. TPU system sales are closer to large infrastructure projects, with high single-transaction amounts, but revenue is easily affected by delivery progress, data center construction, and customer financing. Even if concentrated order confirmations lead to rapid growth, the market may give lower valuations based on hardware business comparisons.

SemiAnalysis predicts that Google Cloud currently has over $150 billion in TPU system backlog orders, and in the coming quarters, it may add over $250 billion in new orders. These figures have not been separately disclosed by Alphabet, and how much revenue can ultimately be converted will depend on whether TPUs can be supplied on time, whether customer data centers can be operational, and whether related financing can be successfully implemented.

Gemini's long-term competitiveness will also affect this business model. In the short term, more computing power sold externally can boost Google Cloud revenue. If Google continues to lag in the field of cutting-edge models, the cloud platform's reliance on third-party models like Anthropic may increase, which could pressure Google's bargaining power and customer stickiness in the model ecosystem.

Next, investors need to closely observe two key metrics: the revenue recognition scale of TPU system orders in 2027 and whether Google Cloud's profit margin can be maintained within the 35% to 39% range predicted by SemiAnalysis.

If orders are delivered smoothly, Google Cloud's revenue growth in 2027 may approach 150%, and Alphabet's AI growth will gain new support. If delivery is delayed or profit margins fall below expectations, this round of growth is more likely to reflect a revenue surge driven by large hardware orders, and the potential for valuation increases will also be limited.