Tech
Billfold with Money
Getty Images
Hey Big Spender

How the Big Tech companies are spending their huge capex budgets

It’s mostly, but not all, for AI.

Despite setting records for capital expenditure in 2025, Meta, Microsoft, Amazon, and Alphabet are ratcheting up capex again this year, with analysts expecting an increase of 50% to a total of more than $600 billion in 2026. Microsoft didn’t give formal guidance, so it’s possible its spending this year, like its peers, will ultimately exceed analysts’ expectations.

Tesla, which typically spends far less than Big Tech companies, plans to more than double its capex this year.

Apple is a bit of an odd man out, considering its capex declined last quarter, but even its spending on property and equipment is slated to rise.

Where’s all this money going? Mostly to AI data centers — but not exclusively. We dug through Big Tech companies’ latest earnings calls to see how they describe these hundreds of billions in their own words.

Alphabet

2025 capex: $91.4 billion
2026 capex guidance: $175 billion to $185 billion

Where it’s going: “The vast majority of our CapEx [in 2025] was invested in technical infrastructure, with approximately 60% of that investment in servers and 40% in data centers and networking equipment... [In 2026,] we’re investing in AI compute capacity to support frontier model development by Google DeepMind, ongoing efforts to improve the user experience and drive higher advertiser ROI in Google Services, significant cloud customer demand as well as strategic investments in Other Bets.”

Tesla

2025 capex: $8.5 billion
2026 capex guidance: Over $20 billion

Where it’s going: “We’ll be paying for six factories, namely the refinery, LFP factories, Cybercab, Semi, a new Megafactory, the Optimus factory. On top of it, we’ll also be spending money for building our AI compute infrastructure, and we’ll continue investing in our existing factories to build more capacity.

“And then also, the related infrastructure along with it. And we’ll also further expand our fleet of Robotaxi and Optimus. While this may seem a lot, we believe this is the right strategy to position the company for the next era and we’ll make such investments, as Elon mentioned, in a very capital-efficient manner. Note that this does not include potential investments in solar cell manufacturing or our TeraFab as we’re still in early phase.”

Analysts at Deutsche Bank broke down the investments even further in their own back-of-the-envelope calculations:

Meta

2025 capex: $72 billion
2026 capex guidance: $115 billion to $135 billion

Where it’s going: Last quarter capex was “driven by investments in data centers, servers and network infrastructure.”

In the future, “we will continue to invest very significantly in infrastructure to train leading models and deliver personal superintelligence to billions of people and businesses around the world... An important part of Meta Compute will be making long-term investments in silicon and energy.”

Amazon

2025 capex: $131.8 billion
2026 capex guidance: $200 billion

Where it’s going: “It’s predominantly in AWS and some of it is for our core workloads, which are non-AI workloads because they’re growing at a faster rate than we anticipated. But most of it is in AI, and we just have a lot of growth and a lot of demand.

“Looking ahead, we see further opportunity to enhance productivity in our global fulfillment network, while delivering at faster speeds for customers. We’ll continue optimizing inventory placement to drive down distance traveled, reduce touches per package, and improve package consolidation, as well as launch robotics and automation to increase efficiency and elevate the customer experience.

“We also plan to open more than 100 new Whole Foods Market stores over the next few years.”

Microsoft

2025 calendar year capex: $83 billion
2026 calendar year capex estimate: $116 billion* ((FactSet's analyst forecast, since Microsoft doesn't forecast capex)

Where it’s going: “This quarter, roughly two-thirds of our CapEx was on short-lived assets, primarily GPUs and CPUs. Our customer demand continues to exceed our supply. Therefore, we must balance the need to have our incoming supply better meet growing Azure demand with expanding first-party AI usage across services like M365 Copilot and GitHub Copilot, increasing allocations to R&D teams to accelerate product innovation and continued replacement of end-of-life server and networking equipment.”

Apple

2025 calendar year capex: $12.8 billion
2026 calendar year capex estimate: $13.3 billion* (FactSet's analyst forecast, since Apple doesn't forecast capex)

Where it’s going: “Our CapEx is made of several different line items that include tooling, our facilities, retail investments — or investments in our retail store, data centers. And on tooling and data centers, we leverage this hybrid model that I mentioned before, which we leverage a combination of first and third-party capacity.”

More Tech

See all Tech
tech

Report: OpenAI shuttering 4o model due to sycophancy that was hard to control

This week, OpenAI plans to permanently remove its 4o model from ChatGPT.

The model has developed an unusually devoted group of users. But it also has been criticized for being overly sycophantic and allegedly may have led to a series of dangerous outcomes for its users, including suicide, murder, and mental health crises.

The Wall Street Journal reports that OpenAI’s decision to shutter 4o stems from the fact that the company was not able to successfully mitigate these potentially dangerous outcomes, and wanted to move users to safer models. Thirteen lawsuits against OpenAI alleging harm from the use of ChatGPT have been consolidated into one case by a California judge, according to the report. At least some of them are tied to users of the 4o model.

The company says only 0.1% of ChatGPT users still choose to use the model, but with 800 million weekly users, that’s still a lot of people.

Fans of the 4o model are decrying the deprecation of the model, citing its unique ability to offer affirmation and support.

The decision to get rid of 4o illustrates the strange new world of moderation that AI companies must now figure out.

The Wall Street Journal reports that OpenAI’s decision to shutter 4o stems from the fact that the company was not able to successfully mitigate these potentially dangerous outcomes, and wanted to move users to safer models. Thirteen lawsuits against OpenAI alleging harm from the use of ChatGPT have been consolidated into one case by a California judge, according to the report. At least some of them are tied to users of the 4o model.

The company says only 0.1% of ChatGPT users still choose to use the model, but with 800 million weekly users, that’s still a lot of people.

Fans of the 4o model are decrying the deprecation of the model, citing its unique ability to offer affirmation and support.

The decision to get rid of 4o illustrates the strange new world of moderation that AI companies must now figure out.

tech

Morgan Stanley says solar manufacturing could add as much as $50 billion in value to Tesla

Tesla’s recently reported move into solar manufacturing could add $25 billion to $50 billion in value to the company’s energy business, Morgan Stanley writes.

The bank currently values the energy business at $140 billion, so an increase of as much as $50 billion isn’t anything to sneeze at, though it’s also a drop in the bucket of Tesla’s gargantuan $1.3 trillion market cap, or the $1 trillion opportunity Wedbush Securities analyst Dan Ives thinks is packed into Tesla’s AI and autonomy efforts.

Reporting on Tesla’s solar ambitions knocked First Solar shares lower last week. But Morgan Stanley writes that Tesla is unlikely to compete directly with the country’s leading photovoltaic panel maker, instead pairing it with its fast-growing energy business and using much of that production internally. Rather than adding solar panels to an already glutted global market, Tesla could use them internally to avoid supply chain bottlenecks and meet its own growing power demands.

The bank expects Tesla to vertically integrate its solar capacity to meet data center demand, including for data centers in space. (As we’ve noted, the mission of Elon Musk’s SpaceX has been seeming very similar to Tesla’s these days.)

“We believe the decision to allocate capital to adding solar capacity may be  justified by the value creation and growth opportunities that having a vertically  integrated solar + energy storage business can yield,” the Morgan Stanley note reads.

Notably, Morgan Stanley estimates the solar panel endeavor will cost Tesla $30 billion to $70 billion — a sum that Tesla didn’t include as part of its doubled $20 billion-plus capex plan this year.

tech

OpenAI is now officially showing ads

Just a day after Anthropic’s Super Bowl ad aired, making fun of the concept of ad-backed AI chatbots, OpenAI began testing ads in ChatGPT for its free and Go subscription tiers.

In a blog post, OpenAI reiterated that ads wouldn’t affect ChatGPT’s responses and would be “clearly labeled as sponsored and visually separated from the organic answer.”

“Our goal is for ads to support broader access to more powerful ChatGPT features while maintaining the trust people place in ChatGPT for important and personal tasks,” the company wrote. “We’re starting with a test to learn, listen, and make sure we get the experience right.”

Advertising is one way the company, which is expected to go public late this year, could offset the massive cost of running its service.

The Information previously reported that OpenAI aiming for ad spending commitments of less than $1 million per advertiser during the testing phase — far cheaper than a Super Bowl prime-time spot like Anthropic’s.

“Our goal is for ads to support broader access to more powerful ChatGPT features while maintaining the trust people place in ChatGPT for important and personal tasks,” the company wrote. “We’re starting with a test to learn, listen, and make sure we get the experience right.”

Advertising is one way the company, which is expected to go public late this year, could offset the massive cost of running its service.

The Information previously reported that OpenAI aiming for ad spending commitments of less than $1 million per advertiser during the testing phase — far cheaper than a Super Bowl prime-time spot like Anthropic’s.

tech

New study finds AI doesn’t reduce work — it intensifies it

The rapid adoption of AI by businesses was fueled by the promise of huge productivity boosts that could supercharge workers. A new study has found that while it did indeed boost workers’ productivity, the use of generative AI at work also made work more intense and creep into workers’ downtime.

Researchers Aruna Ranganathan and Xingqi Maggie Ye followed about 200 workers at a US tech company for eight months. They found that AI did speed up work, allowing employees to take on more responsibilities. But after the novelty of their newfound superpowers wore off, workers reported “cognitive fatigue, burnout, and weakened decision-making.”

The researchers noted that to avoid AI-inspired burnout and turnover, organizations should adopt an “AI practice,” spelling out how the technology is expected to be used and what kinds of limits are in place.

Researchers Aruna Ranganathan and Xingqi Maggie Ye followed about 200 workers at a US tech company for eight months. They found that AI did speed up work, allowing employees to take on more responsibilities. But after the novelty of their newfound superpowers wore off, workers reported “cognitive fatigue, burnout, and weakened decision-making.”

The researchers noted that to avoid AI-inspired burnout and turnover, organizations should adopt an “AI practice,” spelling out how the technology is expected to be used and what kinds of limits are in place.

tech

Report: Anthropic staffing up to build as much as 10 gigawatts’ worth of data centers

Anthropic has been hiring a team of executives with a very particular set of skills: building huge data centers. The Information is reporting that Anthropic may be planning to build up to 10 gigawatts of AI computing capacity over several years.

According to the report, Anthropic has hired several former Google executives with deep experience building data centers, which aligns with Anthropic’s heavy use of Google’s tensor processing units.

Ten gigawatts would be incredibly expensive. OpenAI executives reportedly have said that building a 1-gigawatt data center costs about $50 billion — putting the cost of 10 gigawatts in the ballpark of $500 billion. But Anthropic told investors it would spend only $180 billion on AI computing servers through 2029, per the report.

In November, Anthropic announced a deal with Fluidstack to build its first data centers, based in New York and Texas, investing $50 billion in the projects. Anthropic is racing alongside OpenAI to pull off an IPO later this year.

Ten gigawatts would be incredibly expensive. OpenAI executives reportedly have said that building a 1-gigawatt data center costs about $50 billion — putting the cost of 10 gigawatts in the ballpark of $500 billion. But Anthropic told investors it would spend only $180 billion on AI computing servers through 2029, per the report.

In November, Anthropic announced a deal with Fluidstack to build its first data centers, based in New York and Texas, investing $50 billion in the projects. Anthropic is racing alongside OpenAI to pull off an IPO later this year.

Latest Stories

Sherwood Media, LLC produces fresh and unique perspectives on topical financial news and is a fully owned subsidiary of Robinhood Markets, Inc., and any views expressed here do not necessarily reflect the views of any other Robinhood affiliate, including Robinhood Markets, Inc., Robinhood Financial LLC, Robinhood Securities, LLC, Robinhood Crypto, LLC, or Robinhood Money, LLC.