Alphabet Raises $80 Billion in Record AI Infrastructure Play as Berkshire Hathaway Bets Big on the Cloud

Alphabet, Google's parent company, announced plans to raise roughly $80 billion in equity capital to expand its AI infrastructure and global compute capacity — one of the largest financing packages ever tied directly to the AI buildout. Warren Buffett's Berkshire Hathaway agreed to buy $10 billion of the offering, split evenly between the company's Class A and Class C shares.

The move signals that cloud infrastructure spending has left the territory of routine capital expenditure and entered what analysts call a new phase of industrial-scale capacity formation. The funds will flow into data centers, power delivery systems, networking gear, cooling infrastructure, and AI accelerators — everything needed to turn corporate ambition into actual compute.

"Alphabet's $80 billion equity raise is a sign that AI infrastructure has crossed from ordinary cloud expansion into industrial capacity formation," Ihab Osman, an independent strategist specializing in data center and mission-critical infrastructure, told Data Center Knowledge. "Capital is not capacity until it passes through land, power, equipment, construction, commissioning, and operations."

The Biggest Bets in the Cloud

The scale of Alphabet's fundraising dwarfs most annual technology budgets. The company spent roughly $32 billion on capital expenditures in all of 2023. This single raise is more than double that figure — and it lands in a year when Google is on track to double its data center capex, according to Dell'Oro Group research director Baron Fung.

"Google will double its data center capex this year, and likely not slow down next year as well, exerting some cash flow pressures," Fung said. "That new infrastructure likely won't be utilized and generating revenue right away." He pointed out that large purchases tied to future Nvidia Rubin-based infrastructure and upcoming TPU deployments could create a timing mismatch between infrastructure spending and the revenue those investments eventually generate.

Alphabet's $5 billion TPU deal earlier this year already signaled the company's bet on custom silicon. Google's Tensor Processing Units are designed specifically for AI training and inference workloads, giving the company an alternative to Nvidia's dominant H100 and B200 GPU lineup. The TPU strategy allows Google Cloud to differentiate itself from AWS and Azure by offering specialized AI hardware for customers running large-scale model training.

The three cloud hyperscalers are all racing to secure capacity. AWS posted 28% year-over-year revenue growth in its most recent quarter. Azure climbed 40%. Google Cloud surged 63%, per Synergy Research Group. Google Cloud's Q2 revenue hit $24.8 billion, up 82% from a year earlier — an acceleration that reflects the company's aggressive push into AI services through Vertex AI, Gemini models, and enterprise cloud deals.

Across the industry, technology companies, utilities, developers, and infrastructure investors are committing billions of dollars to projects tied to anticipated growth in AI demand. Attention has focused on securing access to power, transmission capacity, cooling systems, and other infrastructure needed to bring new AI facilities online.

Beyond GPUs — the Whole System Bottleneck

Analysts across the industry agree that the constraint on AI growth is no longer just chip supply. The bottlenecks now cut across energized sites, grid interconnection timelines, transformer and switchgear availability, network fabric, cooling systems, and construction execution. Raising capital is merely the first step in a far more intricate process.

Steven Dickens, CEO and principal analyst at HyperFrame Research, noted that even a company with Alphabet's operating cash flow needed to raise equity. "For a cash-generating machine like Google with such strong businesses as ads, search, YouTube and Workspace to still need funds speaks for itself," he said. Dickens said he is not overly concerned by the raise itself, but suggested investors could begin asking tougher questions if similar financing becomes a recurring feature of Google's AI strategy.

The financing package drew a high-profile vote of confidence from Berkshire Hathaway. Its $10 billion investment expands the conglomerate's position in Alphabet and adds a powerful signal to institutional investors that AI infrastructure is a durable long-term bet. The Berkshire purchase covers both Class A and Class C shares, $5 billion each — a rare move for the famously tech-cautious investor.

"What stands out most is that the bottleneck is no longer only GPUs," Osman said. "It is system capacity: energized sites, grid certainty, transformer and switchgear availability, network fabric, cooling, construction execution, and the ability to convert announced demand into deliverable compute."

Alex Cordovil, research director at Dell'Oro Group, framed the larger picture. "The most telling thing here isn't any single raise or cost cut, it's the optionality," he told Data Center Knowledge. "That breadth of options reads as conviction, not strain. These companies are financing for the long term rather than just solving today's capacity crunch, and a willingness to reach for whatever lever fits tells you they expect AI demand to be durable enough to underwrite it."

Data Center Boom Hits Policy Crossroads

Alphabet's raise sits within a broader data center construction boom that is reshaping commercial real estate and electric grids across the United States. Data centers have become the largest segment of US commercial construction, with project spending reaching a seasonally adjusted annual rate of $51 billion, according to newly released US Census Bureau data — surpassing traditional office development for the first time.

That growth is creating friction with communities and regulators. Virginia, the world's largest data center market, passed a first-of-its-kind tax on data center electricity consumption — $0.011 per kilowatt-hour — that budget documents estimate will generate $600 million annually for the state's general fund over the next two years. The tax took effect July 1, 2026.

Voters in Monterey Park, California, passed a ballot measure prohibiting new data centers within city limits with 86% support, creating what appears to be the first voter-enacted municipal ban on data center development in the country. The measure concluded a months-long fight over a proposed 247,000-square-foot facility backed by Australian investment firm HMC StratCap.

The Federal Energy Regulatory Commission issued show-cause orders to the six largest US grid operators, demanding they defend or rewrite interconnection rules for large loads. FERC's orders set deadlines for markets to define processes for handling gigawatt-scale load requests, evaluating co-located generation, assigning upgrade costs, and ensuring sufficient generation is available when projects come online.

In Wisconsin, Oracle asked a state court to overturn utility regulators' decision requiring some hyperscale data center developers to post hundreds of millions in financial security — a legal test of how states allocate the financial risks of powering AI infrastructure. The company argues that regulators exceeded their authority under new tariffs that target data centers. The case follows the Public Service Commission of Wisconsin's approval of changes to a proposed data center tariff from We Energies, requiring large-load customers to cover the full cost of generation and grid infrastructure serving them.

Texas' Public Utility Commission has approved a one-time, systemwide process to evaluate new large loads and load expansions of 75 MW or more. The overhaul replaces a legacy utility-by-utility approach with a centralized evaluation led by ERCOT, a shift prompted by an avalanche of data center and AI-related power requests. Chevron and Microsoft signed a 20-year agreement to supply dedicated power for a planned data center campus near Pecos, Texas — one of the largest pairings of compute infrastructure and on-site generation in the country.

Global Cloud Infrastructure Goes Big

The global buildout shows no signs of slowing. SoftBank Group announced plans to build a 5 GW AI data center in northern France, representing an investment of up to €75 billion ($85 billion). The first phase is expected to deliver 3.1 GW of AI capacity — enough to power roughly 2.5 million homes. The facility will be one of the largest single AI data center campuses ever planned.

AirTrunk, backed by Blackstone, plans to invest $21 billion in a 3 GW data center campus in Maharashtra, India — a sign that the AI infrastructure buildout is extending deep into emerging markets. Meta inked an agreement with Reliance Industries to lease its first AI-enabled data center in India and is partnering with two clean energy providers to back nearly 1 GW of renewable energy.

Google launched a new data center in Horndal, Sweden, to support Search, Google Cloud, and YouTube services in the country. The facility is the company's first in Sweden and part of a broader push into Nordic markets where cheap hydro and wind power offer attractive operational costs.

Microsoft cut the ribbon on a $3.3 billion facility in Mount Pleasant, Wisconsin, the first data center at its AI campus there, and broke ground on another in La Porte, Indiana. Amazon plans a $10 billion data center campus in Montgomery County, Missouri, while Google will invest $15 billion in a separate project nearby — totaling $25 billion in recently announced hyperscale developments for the county alone.

In the Asia-Pacific region, Alibaba Cloud launched a new public cloud region in Johor, Malaysia. STT Global Data Centers said it will expand its Jakarta campus with 360 MW of AI-ready IT capacity, adding to a campus expected to deliver 2.7 GW by 2030. Digital Edge acquired land in Ansan, South Korea, to develop 60 MW of AI-ready capacity. In South Australia, Iren signed a transmission connection agreement for a planned 800 MW data center campus in Bundey.

Alphabet said the equity raise is needed because "the company is experiencing strong demand for its AI solutions and services from enterprises and consumers, at levels that are exceeding the company's available supply." For Google Cloud customers — from startups training foundation models to enterprises deploying AI agents in production — the message is that Alphabet is betting its balance sheet on having enough compute ready when they need it.

Cloud & Edge Computing — AI — Data Centers on Track to Gobble Up a Fifth of US Electricity by 2035