Google’s $185 Billion Infrastructure Bet Hinges on Cloud’s Payback

Alphabet is sprinting in two directions at once. On the public stage at the Google Cloud Next conference in Las Vegas, the company unveiled its eighth-generation custom AI chip and inked multiyear cloud deals with PepsiCo and Mars. Behind the scenes, however, senior Google managers are fretting about the company’s position in AI-powered coding — a market where rival Anthropic is gaining traction, even among Google’s own DeepMind employees.

The tension between external confidence and internal unease sets the stage for a pivotal earnings report on April 29, when Alphabet will release first-quarter 2026 results after the US market close.

A Chip for Inference, Not Just Training

The TPU 8i represents a strategic pivot in Alphabet’s silicon strategy. Unlike its predecessor, the TPU 8t — which can link up to 9,600 units with two petabytes of high-speed memory in a single superpod for model training — the 8i is purpose-built for inference. That means running live AI applications rather than building them.

Google’s engineers promise higher throughput and lower latency for real-time workloads. The logic is straightforward: proprietary chips reduce dependency on third-party hardware, cutting costs for both Alphabet and its cloud customers. The TPU 8t remains the training workhorse, but the specialization signals that Google sees inference as the next battleground for enterprise AI.

Big Consumer Brands Go All-In on Gemini

PepsiCo will deploy the Gemini Enterprise Agent Platform for supply chain management and sales operations, while Mars signed a separate Google Cloud agreement. These are not trial runs — both are multiyear commitments that reflect a broader shift from AI experimentation to full-scale enterprise integration.

The momentum is visible in the numbers. Google’s models now process more than 16 billion tokens per minute through the direct API, up from 10 billion in the prior quarter. Paid monthly users of Gemini Enterprise jumped 40% quarter-over-quarter in Q1. Internally, AI now generates 75% of all new code at Google, compared with 50% last autumn.

Should investors sell immediately? Or is it worth buying Alphabet?

The Coding Anxiety Beneath the Surface

That last statistic masks a deeper concern. According to Bloomberg, leading Google managers are worried about the company’s competitive position in AI-assisted developer tools. Anthropic’s Claude model is winning enterprise customers — and reportedly, some DeepMind staff are using it despite internal policies against external tools.

The structural problem is fragmentation. Google’s coding capabilities are scattered across more than six products under different brand names, diluting their collective impact. In response, DeepMind has assembled a specialist team under engineer Sebastian Borgeaud focused on complex, long-horizon programming tasks like writing new software from scratch. Co-founder Sergey Brin and DeepMind CTO Koray Kavukcuoglu are directly involved. Brin’s internal memo was blunt: “The final battlefield in the AI competition is agents.”

The $185 Billion Question

Alphabet has earmarked capital expenditures of $175 billion to $185 billion for full-year 2026 — nearly double the $91.4 billion spent in 2025. The money is flowing into AI infrastructure to service a growing cloud order backlog, but the depreciation from new data centers is squeezing margins. Some analysts project negative free cash flow for 2026.

The analyst consensus calls for earnings per share of $2.68 on revenue of roughly $107 billion for the first quarter. BMO Capital raised its price target to $410 from $400 with an Outperform rating, while JPMorgan reaffirmed its Overweight rating, calling Alphabet a “Top Pick” as the Las Vegas conference kicked off. Morningstar rates the stock three out of five stars as fairly valued, with a $340 target.

The stock closed Wednesday at $339.32, up 2.1%, and trades around €291 — just shy of its 52-week high. Since the start of the year, shares have gained roughly 8%.

All eyes will be on Google Cloud’s revenue growth when Alphabet reports on April 29. Some analysts expect a year-over-year increase of more than 50%. But the critical question for the earnings call will be whether the pace of spending is starting to dent margins — and whether the investment program is leaving measurable traces in cloud revenue momentum.