Alphabet raises annual capital expenditure guidance to USD 205 billion, massively increasing investment in AI data centers, custom silicon, and networking infrastructure.

Alphabet has supercharged the artificial intelligence infrastructure race, raising its full-year 2026 capital expenditure forecast to as much as $205 billion. The move, disclosed during the company's second-quarter earnings release on July 22, signals that the hyperscaler is betting heavily on cloud computing and custom silicon to capture a larger slice of the AI market, sending a clear demand signal to chip suppliers Nvidia, Broadcom, and Micron Technology.

CEO Sundar Pichai told analysts that the dynamics around AI infrastructure "look healthier than where we were about a year ago," as Alphabet lifted its CapEx range from a previous high of $190 billion to a new band of $195 billion to $205 billion. The decision comes as Google Cloud revenue surged 82% year over year to $24.77 billion, with an operating margin expanding to 35.6% and a staggering $514 billion backlog. Nearly 90% of the Fortune 100 now use Gemini Enterprise, and the Gemini App has reached 950 million monthly active users.

Alphabet's move lands in the middle of a broader escalation: all four U.S. hyperscalers lifted or reaffirmed elevated 2026 capital spending plans within the same earnings season. Combined, the four companies are tracking toward roughly $725 billion in 2026 capital spending, about 77% above 2025 levels.

Alphabet's spending spree directly benefits Nvidia, whose graphics processing units remain the gold standard for accelerated computing. Alphabet has long been a major Nvidia customer, using its GPUs for internal workloads and renting them out through Google Cloud. But the company is also deepening its partnership with Broadcom to develop custom AI chips, known as tensor processing units, that it began selling directly to external clients in the second quarter. Broadcom, which designs custom XPUs for Google, Meta Platforms, OpenAI, and Anthropic, reported AI semiconductor revenue of $10.8 billion in its latest quarter—a 143% jump—and guided for $56 billion in full-year AI semi revenue, with AI bookings already exceeding $30 billion.

Micron stands to benefit from the memory requirements of massive data center expansions. Each new server cluster demands high-bandwidth memory, a segment where Micron is a key supplier alongside South Korean rivals. Analysts see the elevated CapEx guide as confirmation that the AI buildout cycle has years to run, with hyperscaler spending likely to remain elevated through 2027 and beyond.

The spending surge carries financial strain. Alphabet's second-quarter free cash flow swung to negative $5.86 billion, a 210% year-over-year deterioration, as quarterly CapEx hit $44.9 billion. Long-term debt nearly doubled to $98.2 billion, the share buyback program was suspended, and interest expense rose almost fivefold. Headline net income of $112.1 billion was inflated by $99 billion in unrealized equity gains, introducing mark-to-market volatility. Still, with Google Cloud's backlog expected to convert over 50% into revenue within 24 months, the cash flow drought could reverse if margins hold.

Wall Street remains broadly constructive. Alphabet trades at roughly 17 times forward earnings against 24% top-line growth, with a consensus analyst target of $428.04, implying about 18% upside from recent levels around $362. The stock has 58 buy ratings against six holds and zero sells, and has outpaced the S&P 500 since the Q2 filing.

Alphabet's custom silicon strategy is emerging as a structural differentiator. The company deployed its first TPU in 2016 and made the chips available to cloud customers in 2018. In the latest quarter, it began selling TPUs directly to clients for use in external data centers, positioning itself as a more direct competitor to Nvidia. Gil Luria, head of technology research at D.A. Davidson, estimates Alphabet could eventually capture 20% of the AI infrastructure market, valuing its chip business at around $900 billion. Morgan Stanley analysts project custom silicon, primarily Alphabet's TPUs, will account for 24% of AI accelerator sales by 2030, up from 15% today.

The AI accelerator market is projected to top $300 billion this year and could reach $600 billion by 2030, according to industry estimates. That would translate into more than $100 billion in annual TPU revenue for Alphabet by the end of the decade if current trajectories hold. Anthropic and Meta have already committed billions of dollars to TPU purchases, and Alphabet recently announced a joint venture with Blackstone to build a TPU cloud business. The broader AI-ready data center buildout—chips, power, cooling, and construction combined—will require $5.2 trillion in cumulative capital spending through 2030, according to McKinsey, underscoring how much of the current hyperscaler capex wave still has to play out.

Nvidia is unlikely to cede its dominance quickly. The company's proprietary CUDA software platform, comprising hundreds of code libraries and frameworks, creates a durable economic moat that makes switching prohibitively expensive for developers. GPUs also run a far broader set of algorithms than TPUs, which are optimized for specific deep learning workloads. Once a team has built pipelines on CUDA, switching to another platform is prohibitively expensive. Still, the hyperscaler shift toward custom silicon is unmistakable. Amazon and Microsoft have also deployed purpose-built AI chips, but Alphabet's decade-long head start makes it the most significant threat to Nvidia's market share.

Broadcom occupies a unique position in this landscape. Its networking business supplies the Ethernet fabric that stitches together clusters of 100,000 GPUs, meaning it collects a toll on every hyperscaler build regardless of whether the compute comes from Nvidia, AMD, or in-house silicon. CEO Hock Tan guided third-quarter AI revenue to $16 billion, a jump of more than 200%, with visibility extending into 2028.

For investors, the CapEx raise reinforces a thesis that AI infrastructure spending is still in its early innings. Nvidia, Broadcom, and Micron each offer differentiated exposure to the buildout. Nvidia remains the compute standard, Broadcom captures custom silicon and networking, and Micron supplies the memory backbone. The primary risk is that cloud growth decelerates or margin compression from third-party capacity delays the return to positive free cash flow. Polymarket assigns an 82.2% probability to a Gemini Pro release by August 31 and a 93.5% chance the next model debuts with strong performance scores, near-term catalysts that could sustain Alphabet's momentum.

Alphabet's decision to push CapEx toward $205 billion is a bet that the AI opportunity is large enough to justify the balance sheet strain. With a $4.35 trillion market cap, it is the largest AI hyperscaler, and its spending decisions ripple across the entire semiconductor supply chain. For chip investors, the message is clear: the AI capex supercycle is far from over.