Google CEO Sundar Pichai gestures during a meeting with French President Emmanuel Macron on the sidelines of the AI Impact Summit in New Delhi on February 19, 2026.
Ludovic Marin | AFP | Getty Images
Wall Street analysts estimate that total AI capital spending could exceed $1 trillion by 2027, following even larger spending plans announced by hyperscalers in Wednesday’s tech results.
Both Evercore and Bank of America expect capital spending to exceed $1 trillion in 2027 after reporting their results, with estimates for 2026 rising to between $800 billion and $900 billion.
“Capital spending continues to rise as demand exceeds supply and prices are rising,” Jefferies analysts said in a note to investors Thursday.
Spending forecasts for this year rise across the board, including parent company Google alphabet 4% increase to $185 billion; Amazon 1% increase to $200 billion; meta 8% increase to $135 billion; microsoft Sales rose 24% to $190 billion, according to figures compiled by Bank of America.
Tech company CEOs are expressing confidence in their companies’ investments in artificial intelligence as evidence of monetization, such as increased cloud revenue, is reflected in the latest earnings reports, but rising spending continues to generate skepticism among investors.
Amazon CEO Andy Jassy said the company is “confident in long-term capital spending” and plans to invest $200 billion annually.
Alphabet’s cloud revenue rose 63% in the first quarter from a year earlier, pushing its stock price up about 10%. Chief Financial Officer Anat Ashkenazi said Wednesday that capital spending plans are increasing to meet “solid demand.”
investors want returns
While the overall cost of building AI is worrying, analysts say they are seeing a flow-through from investment to revenue as valuations and market capitalizations soar.
Analysts at Jefferies said: “While capital spending continues to rise, the ROI is evident with an approximately $2 trillion backlog and accelerating cloud growth.” “Despite AI investments, hyperscalers’ margin leverage has been maintained, highlighting structural (operating expense or) operating expense discipline.”
Confidence in monetization is particularly high for Alphabet, whose growing backlog supports its computing inventory and expansion.
“The backlog confirms a capital spending supercycle,” Brian Pitts wrote Thursday at BMO Capital Markets. “Google’s backlog nearly doubled quarter-over-quarter and grew 400% annually to $462 billion.”
“The bulk of the backlog is core Google Cloud Platform contracts, of which Google expects to recognize just over 50% as revenue over the next 24 months,” he added.
While Google’s cloud revenue has impressed analysts, Meta’s expansion plans have troubled investors who were hoping for a better return on the investments the company has made. The stock price has recently fallen about 8%.
“Meta is likely to remain in the penalty box until capital investment ROI becomes more clear,” Jefferies analysts wrote in a note Thursday.
The company spent $72 billion on capital expenditures in 2025, and expects that to double to $125 billion to $145 billion in 2026. This is up from the previous range of $115 billion to $135 billion.
Meta CEO Mark Zuckerberg said Wednesday that “we are increasing our infrastructure capital spending forecast for this year.” “Most of that is due to rising component costs, especially memory. But all the signs we’re seeing in our own work and across the industry give us confidence in this investment.”
Meta’s free cash flow has been declining, falling to just $1.2 billion in the first quarter from $26 billion a year ago.
Analysts at Bank of America said they expect sales and free cash flow to improve across the sector in 2026, supporting spending.
Who benefits?
Sustained increases in capital spending are good news for chipmakers and gear providers. They are a hyperscaler customer, and analysts were keeping an eye on them Thursday. First quarter profits of CPU manufacturers intel It was especially powerful because AI buildouts require more than just graphics processing units (GPUs).
“Demand for a variety of custom (application-specific integrated circuit) programs (TPU, Trainium, Maia, MTIA) is strong and growing,” according to Evercore analysts, touting “a focus on agent AI as a key use case, which we believe will prove to drive a CPU renaissance in the coming years.”
RBC Capital Markets maintains positive rating Nvidia, micron technology, marvel, Astera Research Institute, arm holdingsand lattice semiconductor.
“The solid trend in capital spending should also bode well for AVGO, AMD, SNDK and INTC, which received sector performance reviews,” RBC said. “Wafer fabrication fabs are experiencing double-digit growth due to the demand for AI.”
