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

TL;DR
·SemiAnalysis believes that following the leadership changes at DeepMind, Google Cloud could be the biggest short-term beneficiary.
·Alphabet's Google Cloud revenue in Q2 grew by 82% year-over-year to $24.768 billion, with a cloud business backlog of $514 billion.
·The report estimates that orders for over $150 billion in TPU systems could accelerate Google Cloud's revenue growth to around 150% by 2027.
·This forecast depends on order deliveries, customer financing, and revenue recognition timing, with a valuation of system sales potentially lower than traditional cloud services.

On August 7th, SemiAnalysis released a report stating that following the leadership changes at Google DeepMind, Google Cloud could become the largest short-term financial beneficiary.

The report connects the personnel changes with Google's internal allocation of computing power. TPUs are scarce resources that both Google's cutting-edge models and cloud business rely on. More computing power used for Gemini training will help Google catch up with cutting-edge models; more TPUs flowing to external customers will lead to faster conversion into Google Cloud revenue, backlog, and profit.

Based on this, SemiAnalysis makes a set of aggressive predictions: TPU system sales could accelerate Google Cloud's revenue growth to around 150% by 2027, significantly higher than the report's listed 64% seller consensus expectations, contributing approximately $3 to Alphabet's EPS for that year.

DeepMind Leadership Change Prompts Market to Reassess Computing Power Allocation

According to Axios, Demis Hassabis stepped down as Google DeepMind CEO, assuming the role of DeepMind Chairman and Alphabet's Chief Scientist; former DeepMind CTO Koray Kavukcuoglu took over day-to-day management, reporting to Alphabet CEO Sundar Pichai.

Simultaneously, Jeff Dean, Sanjay Ghemawat, Oriol Vinyals, and Quoc Le left Google to found the AI research institution Discovery Loop. Google will participate as a founding investor and establish cloud service partnerships with them.

This round of adjustments has prompted a reevaluation of Google's computing power allocation strategy in the market. The same batch of TPUs is required to support both Gemini training and inference, be made available to external AI labs and enterprise customers through Google Cloud, and be sold as a whole to specialized AI data center operators.

Different allocation methods correspond to different rewards. Investing computing power in Gemini is related to Google's competitiveness in cutting-edge model development, while TPUs allocated to commercial customers can accelerate Google Cloud revenue and backlogged orders.

SemiAnalysis assesses that Google Cloud leader Thomas Kurian's influence in internal computing power competition is on the rise. With more TPUs and complementary infrastructure flowing to external customers, Google Cloud may gain greater revenue elasticity.

DeepMind's share of Google AI computing power declines

However, Alphabet did not attribute this organizational adjustment to delays in Gemini development. The company stated in its second-quarter earnings call that Gemini 4 is undergoing its most ambitious pre-training to date, with AI services still facing supply constraints. Pichai also emphasized that the primary goal of TPU allocation remains to ensure that Google has the computing power necessary to participate in the cutting-edge AGI competition.

Currently, Google has not announced a reduction in the priority of Gemini's computing power. "Gemini Priority Reduction" is SemiAnalysis' inference regarding personnel changes and resource allocation, rather than a confirmed strategic adjustment by Google.

Cloud Revenue Surges by 82%, TPU System Sales Begin to Reshape Growth Structure

Alphabet's second-quarter earnings report shows that Google Cloud revenue grew by 82% year-over-year to $24.768 billion, with cloud business backlog reaching $514 billion. Google's proprietary Model API processed approximately 22 billion tokens per minute, up from 16 billion in the previous quarter.

Google Cloud's rapid growth is not solely from traditional cloud services. Alphabet has begun delivering complete TPU systems to customer data centers and recognized related revenue for the first time in the second quarter. The company stated that TPU system sales contracts have been included in the cloud business backlog, with a significant portion of the revenue from existing contracts expected to be recognized by 2027.

This signifies a change in Google Cloud's revenue structure. In the past, cloud revenue mainly came from recurring services such as compute, storage, databases, and AI platforms; now, sales of complete TPU systems are starting to become a new growth driver.

Percentage of TPU Direct Sales to Anthropic

According to SemiAnalysis, revenue from TPU system sales, recognized on a total basis, is estimated to be approximately $35 billion per GW, with related revenue recognized in the second quarter of 2026 totaling around $1.2 billion. Based on their analysis, excluding this revenue, Google Cloud's core business's year-over-year growth rate is still at the forefront at over 70%; including TPU system sales, the overall growth rate rises to 82%.

The two types of revenue have different economic characteristics. Traditional cloud business depends mainly on customer renewals, computing usage, and platform stickiness; TPU system sales are more akin to large AI infrastructure projects, which, once delivered, could significantly boost quarterly or annual revenue growth.

A $150 Billion Order Pending Confirmation, 2027 Growth Rate May Surge to 150%

One of the most impactful forecasts from SemiAnalysis stems from the backlog of TPU system orders.

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

Google Cloud Revenue Growth Rate Forecast

The $514 billion cloud business backlog disclosed by Alphabet includes TPU system sales, but the scale of TPU orders was not separately disclosed. Therefore, the figure exceeding $150 billion comes from SemiAnalysis's modeled estimation.

Based on the report's adoption of a revenue recognition criterion of around $35 billion/GW, a single GW-scale project is enough to significantly alter Google Cloud's growth trajectory. SemiAnalysis also predicts that as AI labs continue to procure large-scale computing power, there may be more multi-GW transactions in the coming quarters, adding over $250 billion in TPU orders to Google Cloud's remaining performance obligations.

The profit side may also receive support. SemiAnalysis expects the EBIT profit margin from TPU system sales to be slightly below the core cloud business's level of over 30%, but Google Cloud's overall EBIT profit margin may still remain at 35% to 39%. With high revenue growth and a higher profit margin working together, the related business may contribute approximately $3 EPS to Alphabet in 2027.

As a result, SemiAnalysis has linked a personnel adjustment in the AI division to a complete financial chain: the Gemini organizational change affects computing power allocation, more TPUs flow to external customers, system sales expand cloud business backlog orders, and ultimately contribute to Google Cloud revenue and Alphabet's profit.

Even with Gemini Lagging, Google Can Still Profit from the AI Race

SemiAnalysis takes a notably pessimistic view of Gemini's long-term competitiveness, but for investors, the more easily verifiable issues are: who ultimately uses the expensive TPU clusters, who pays for them, and whether these orders can translate into Google Cloud revenue.

The report notes that Gemini's first-party API token growth has slowed down. In the first quarter of 2026, its throughput increased from 100 billion per minute to 160 billion, a 60% increase quarter-over-quarter; further increasing to 220 billion in the second quarter but with a growth rate dropping to about 38%.

This does not mean that Gemini's usage has declined, but rather its growth rate has slowed. Google's official statement still indicates strong demand for the model API, with the company facing computational supply constraints, and Gemini 4 pre-training is still progressing.

Gemini First-Party API Revenue and Growth

At the same time, Google's enterprise AI platform does not only host Gemini but also offers third-party models like Claude to customers. Even if Google's in-house models are no longer consistently leading, Google Cloud still has the opportunity to capture the entire AI industry's capital expenditure through selling computing power, platform services, and TPU systems.

The Discovery Loop may also enter this cycle. Google is not only an initial investor in this entity but will also establish cloud service cooperation with it. SemiAnalysis further speculates that researchers leaving Google may receive financing from external investors and Google-related capital to then use some of the funds to purchase computing resources on Google Cloud.

This leads to a counterintuitive business outcome: even if Google is under pressure in cutting-edge model competition, it may still rely on cloud infrastructure to generate revenue from external AI investments.

Investor Metric: Despite 150% Growth Rate, TPU Order Value Still Needs Validation

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

Traditional cloud businesses rely on customer renewals and usage growth, leading to relatively stable revenue, while the cost of platform migration also enhances customer stickiness. TPU system sales are more akin to large infrastructure projects, with high individual amounts, yet revenue can easily be affected by delivery timelines, data center construction, and customer financing. Even with rapidly growing confirmed orders, the market may assign a lower valuation based on hardware business comparisons.

SemiAnalysis predicts that Google Cloud currently holds over $150 billion in TPU system backlog orders, with the potential for an additional $250 billion in orders over the next few quarters. These figures have not been separately disclosed by Alphabet. The ultimate conversion of these orders into revenue will depend on TPU's ability to supply on time, customer data center readiness, and the smooth execution of related financing.

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

Going forward, investors need to focus on two key metrics: the scale of income recognition from TPU system orders in 2027 and whether Google Cloud's profit margin can maintain the 35% to 39% range as predicted by SemiAnalysis.

After successful order deliveries, Google Cloud's revenue growth in 2027 could approach 150%, and Alphabet's AI growth will also find new support. If deliveries are delayed or profit margins fall below expectations, this growth is more likely to reflect a revenue surge driven by large hardware orders, limiting potential valuation upside.