Big Tech's $700 Billion AI Buildout: What MSFT, GOOGL, META, AMZN, and AAPL Are Spending in 2026

Q1 2026 earnings calls turned the AI capital expenditure race into a numbered game. Within 48 hours, Microsoft, Alphabet, Meta, and Amazon all disclosed full-year capex plans, and three of the four raised them. Apple reported the next day with an entirely different number on a different scale. Combined, the five companies plan to spend more than $700 billion on capital projects in 2026, the bulk of it on AI infrastructure.

The headline numbers are easy to compare. The strategic stories behind them are not. Each company is buying GPUs, data centers, and power, but each frames the spending around a different end product: enterprise cloud, search and ads, consumer superintelligence, retail-plus-AWS, or on-device intelligence. Below is what each disclosed in Q1 2026, and what management said the money is for.

The Numbers, Side by Side

Microsoft: $190 Billion, Mostly GPUs

Microsoft spent $31.9 billion on capital expenditures and finance leases in fiscal Q3 2026, up 49% year over year. CFO Amy Hood disclosed that calendar 2026 capex will reach roughly $190 billion, including about $25 billion of incremental spend driven by higher component prices. Next quarter alone is guided above $40 billion.

Hood specified that roughly two-thirds of the capex is going to short-lived assets, primarily GPUs and CPUs. The remaining third is data center shells, networking, and long-lived real estate. The mix is the closest disclosure any hyperscaler has provided on the silicon-versus-concrete split, and it implies Microsoft is buying close to $125 billion of compute hardware in 2026 alone.

The justification is the AI revenue ramp. Microsoft's AI business now runs at a $37 billion annual revenue rate, up 123% year over year, and Azure grew about 40% in the quarter. Intelligent Cloud revenue reached $34.7 billion, up 30%. Management framed the capex as demand-driven: customer commitments are filling capacity faster than it can be brought online.

Alphabet: $190 Billion and "Compute Constrained"

Alphabet reported $35.7 billion in Q1 2026 capex covering real estate, servers, data centers, and other infrastructure. CFO Anat Ashkenazi raised the full-year 2026 capex range to $180-$190 billion, up from $175-$185 billion. She also told analysts the 2027 number will "significantly increase" from the 2026 baseline, the most explicit forward signal any hyperscaler gave on the call.

CEO Sundar Pichai explained the raise plainly: "We are compute constrained in the near term." Google Cloud reached 18% of total Alphabet revenue in Q1 2026 and grew at the fastest rate of any segment. Search and YouTube ad revenue both posted double-digit growth, with Pichai attributing part of the lift to AI Overviews and Gemini-powered search experiences. The capex is therefore funding two demand vectors at once: external Google Cloud customers and internal AI-driven product features.

Investors rewarded the spending plan more than they did Microsoft's or Meta's. Alphabet shares rose on the print, while Microsoft and Meta both traded lower in extended hours despite beats. The market read Alphabet's capex as more clearly tied to monetizing assets that already exist.

Meta: $145 Billion for "Personal Superintelligence"

Meta delivered the quarter's highest revenue growth rate, $56.31 billion in revenue up 33% year over year, the company's fastest growth pace since 2021. It also delivered the quarter's largest negative stock reaction. Shares fell about 7% in extended trading after CFO Susan Li raised full-year capex guidance to $125-$145 billion, up from $115-$135 billion. Full-year operating expenses were guided to $162-$169 billion.

Meta's capex framing differs from the cloud players. Meta does not sell its compute capacity to external customers. Every dollar of GPU spend is justified internally by improvements to ranking, recommendations, advertising, and the long-arc bet on consumer AI products. CEO Mark Zuckerberg used the call to introduce "Meta Superintelligence Labs" and stated the company is "on track to deliver personal superintelligence to billions of people."

The market reaction reflected the asymmetry. Microsoft, Alphabet, and Amazon can point to enterprise customers signing multi-year capacity contracts. Meta is asking shareholders to fund a research bet whose return is harder to underwrite quarter by quarter. The 33% revenue growth shows the existing AI investments in Reels and ads are working; the additional $10 billion of 2026 capex is funding the next bet beyond that.

Amazon: $200 Billion, and a 95% Drop in Free Cash Flow

Amazon reported $181.5 billion in revenue, up 17%, with AWS growing 28% year over year to $37.59 billion, the cloud segment's fastest growth in more than three years. Cash capex in Q1 was $43.2 billion, primarily directed at AWS and generative AI. The full-year 2026 capex plan, first disclosed in February, stands at roughly $200 billion.

The clearest cost of the buildout shows up in free cash flow. Trailing-twelve-month free cash flow fell to $1.2 billion, a 95% decrease year over year, almost entirely because of AI investment. CEO Andy Jassy framed the spending as a structural lead-lag problem: AWS lays out cash for land, power, buildings, and hardware "6 to 24 months before we start billing customers." He stated Amazon has "high confidence" the 2026 capex will be "monetized well," with "customer commitments for a substantial portion of it."

Of the four hyperscalers, Amazon's capex is the most directly tied to a single business unit. AWS is the customer for most of the spend, and AWS revenue is what has to grow to justify it. The 28% acceleration is the case for the spending; the 95% free cash flow decline is the case against. Both are real.

Apple: $14 Billion, On-Device, By Choice

Apple reported fiscal Q2 2026 the day after the four hyperscalers, with $111.2 billion in revenue, up 17%, and net income of $29.6 billion. The capex disclosure was the day's most striking contrast. Through the first half of fiscal 2026, Apple has spent roughly $4.3 billion on capex. Full-year capex is expected to come in around $14 billion. The four hyperscalers will spend more than $650 billion combined.

Apple is not declining to invest in AI; it is investing through a different line item. R&D rose 33% year over year to $11.42 billion in the quarter, faster than revenue growth. The architectural choice that explains the capex gap is on-device inference: Apple Intelligence runs primarily on Apple Silicon (A-series and M-series chips) inside the user's iPhone, iPad, or Mac, not in a hyperscale data center. The cost of that compute sits inside the device bill of materials, not on Apple's balance sheet as data center capex.

Where Apple needs cloud-scale models, it is increasingly partnering rather than building. Earlier in the quarter Apple announced it would use Google's Gemini to power the next generation of Siri's reasoning capabilities. Tim Cook on the call said Apple is "clearly investing more" in AI but described it as incremental on top of the existing product roadmap, not a strategic pivot. The board also authorized an additional $100 billion in stock buybacks, the clearest signal that Apple does not believe it needs to redeploy that cash into AI infrastructure.

What the Money Is Actually Buying

Across the four hyperscalers, the dollars route to four categories: GPUs (NVIDIA H100/H200/B200, plus custom silicon like Google TPU, AWS Trainium/Inferentia, Meta MTIA), CPUs and networking, data center shells and land, and power. Microsoft's two-thirds-on-silicon disclosure suggests the chip share is dominant for cloud-first players. Meta's MTIA program and Google's TPU program are partially capex offsets, designed to reduce the share of every dollar that flows to NVIDIA over time.

Two factors are inflating the 2026 numbers beyond the underlying chip count. First, component pricing. Both Microsoft and Meta cited higher component prices as a contributor to the 2026 raise. Roughly $25 billion of Microsoft's $190 billion 2026 plan is incremental cost from price, not incremental capacity. Second, power. Hyperscalers are signing multi-decade power purchase agreements and, in some cases, financing nuclear restarts to lock in capacity. Those commitments show up across capex and operating cost lines and increasingly drive the location of new data centers.

The aggregate matters for the broader market. Combined 2026 capex from the five companies sits in the $700-$725 billion range, with hyperscaler capex alone above $650 billion. That spending flows through to NVIDIA, AMD, Broadcom, TSMC, Vertiv, Eaton, and the construction and power supply chains. The earnings season did not just reset expectations for Microsoft, Google, Meta, and Amazon; it reset the demand outlook for everything they buy.

Side-by-Side Summary

• Microsoft: ~$190B calendar 2026 capex, +49% Q3, ~2/3 GPUs and CPUs. AI run rate $37B, up 123%.
• Alphabet: $180-$190B 2026 capex (raised), $35.7B in Q1, 2027 to "significantly increase." Pichai: "compute constrained."
• Meta: $125-$145B 2026 capex (raised). Revenue +33%. Funds internal AI products and Meta Superintelligence Labs.
• Amazon: ~$200B 2026 capex. AWS +28% (3-year high). Free cash flow down 95% YoY to $1.2B TTM.
• Apple: ~$14B fiscal 2026 capex. R&D +33% to $11.4B. On-device AI strategy, partnership with Google Gemini for Siri.
• Combined hyperscaler capex (MSFT + GOOGL + META + AMZN): more than $650B in 2026.
• Apple's full-year capex equals about 2.2% of the hyperscaler total.

What to Watch in Q2 2026

Three numbers will tell the story of whether the 2026 capex plans hold. First, Microsoft's next-quarter capex guide of "over $40 billion". Any actual number meaningfully above that signals further demand pull-in. Second, Alphabet's Google Cloud growth rate, which has to keep accelerating to absorb the raised capex. Third, Amazon's free cash flow trajectory, which cannot stay near zero indefinitely without raising questions about the AWS commitment-coverage Jassy described.

For the AI thesis as a whole, the asymmetry between Apple's $14 billion and the hyperscalers' $650 billion is the trade. Either centralized cloud AI continues to capture the bulk of value, justifying the hyperscaler spend, or on-device intelligence captures more of the consumer surface than expected, validating Apple's restraint. Q1 2026 earnings did not resolve that question. They sharpened it.

Beta Finch covers each of these earnings calls in full. The episodes embedded above and linked below break down the numbers, the management commentary, and the strategic context for each company in detail.