Alphabet's $190 Billion AI Bet: A SpaceX GPU Lease, a $10 Billion Berkshire Stake, and a New Open-Source Model

Alphabet raises $85B, secures $30B SpaceX GPU lease, unveils faster AI model, but stock slips 2% as market digests massive capex plans.

Alphabet is pulling off a financial and technological hat trick that would stagger most companies — raising $85 billion in fresh capital, locking in a $30 billion GPU lease with SpaceX, and rolling out a next-generation open-source AI model that runs four times faster than conventional alternatives. The market, however, is not yet applauding: the stock slipped 2.17% on Wednesday to €309.10, widening the gap from its May all-time high.

The centerpiece of the funding blitz is a near-record $85 billion equity package, of which $19 billion comes in the form of mandatory convertible notes — traded under the tickers GOOGM and GOOGN — carrying a 6.25% dividend. Joining the round is Warren Buffett’s Berkshire Hathaway, which is reportedly chipping in $10 billion. The proceeds are earmarked for an extraordinary capital expenditure programme: Alphabet plans to spend between $180 billion and $190 billion in 2026 alone, almost entirely on building out data-centre infrastructure for artificial intelligence.

To power that infrastructure, Alphabet has struck a massive deal with SpaceX. Starting at the end of 2026 and running through mid-2029, Google will pay roughly $920 million per month for access to about 110,000 NVIDIA graphics processing units. Analysts estimate the contract’s total value at around $30 billion. The arrangement comes just ahead of SpaceX’s expected initial public offering, and Alphabet already holds a substantial stake in the rocket company — a position that, after the IPO, could be worth more than $100 billion, providing the tech giant with a significant liquidity cushion.

On the product side, Google DeepMind has unveiled DiffusionGemma, an open-source model that processes data in parallel blocks rather than sequentially, delivering text generation at speeds exceeding 1,000 tokens per second on Nvidia H100 chips. While its quality does not yet match the flagship Gemma 4, the speed advantage makes it ideal for real-time local applications. Separately, Alphabet launched Gemini 3.5 Live Translate, a tool that translates speech directly into more than 70 languages without the traditional transcription pipeline, preserving tone and rhythm. Google is integrating the technology into Meet and the Translate app; Southeast Asian ride-hailing giant Grab is already using it for roughly 10 million voice calls per month. At the same time, Alphabet cut the price of its “Google AI Plus” subscription from $7.99 to $4.99 while doubling the included storage to 400 GB.

The sheer scale of the investment is beginning to show in the numbers. In the first quarter of 2026, Google processed 16 billion tokens per minute — a 60% quarter-on-quarter increase — while revenue climbed 22% year over year. BMO Capital ranked Alphabet first in its “Full Stack AI Score,” citing leadership in infrastructure, models, and product integration. TD Cowen reiterated its Buy rating and a $475 price target, convinced the spending will be justified by accelerated cloud growth.

Yet the stock has felt pressure. At €309, it sits just above its 50-day moving average but remains nearly 12% below the 52-week high of €350.75. The relative strength index stands at around 41, approaching the oversold threshold. Still, the shares are trading almost 17% above the 200-day average of €264.10, keeping the long-term uptrend intact. Over the past twelve months, Alphabet has delivered a gain of roughly 98%, and year-to-date it remains up about 15%.

A minor legal setback in Los Angeles — where a judge rejected a motion by Google and Meta for a new trial over platform responsibility for young users — had little impact on the broader narrative. The real test for Alphabet is whether its $85 billion war chest and $30 billion GPU lease can translate into the revenue growth needed to justify the most ambitious infrastructure build-out in corporate history.