Microsoft vs Amazon AI Earnings: Copilot Monetization, AWS Growth and Soaring Capex

The figures below cover Microsoft's fiscal fourth quarter, which ended June 30, 2026, and Amazon's second quarter, which ended the same day. They are not current-quarter data.

The headline numbers​

Microsoft reported on July 29 and Amazon on July 30.

Microsoft (quarter ended June 30, 2026)

• Revenue was $90.0 billion, up 18%.
• Operating income was $40.6 billion, up 18%.
• GAAP diluted EPS was $4.81. Non-GAAP EPS was $4.74, which excludes the impact of Microsoft's OpenAI investments.
• Microsoft Cloud revenue was $59.3 billion, up 27%.
• Azure and other cloud services grew 43%.
• Commercial remaining performance obligation (RPO) rose 84% to $678 billion.

Amazon (quarter ended June 30, 2026)

• Net sales were $200.6 billion, up 20% from $167.7 billion a year earlier.
• Operating income rose 43% to $27.5 billion.
• AWS operating income reached $16.6 billion, up from $10.2 billion, at a 39.4% operating margin.
• AWS revenue of $42.2 billion was up 36.7%, the fastest growth in 18 quarters. The annualized run rate is $169 billion.

Read the EPS lines carefully​

Both companies booked gains on Anthropic. The scale differs enormously.

Microsoft said its quarter included a $3.2 billion gain from its Anthropic investment. That, together with other discrete items, gave a $0.27 per-share benefit against its prior guidance. Microsoft also said that adjusted for those items it still beat expectations on revenue, operating income and EPS.

Amazon's gain was far larger. Its second-quarter net income included $53.4 billion of non-operating pre-tax other income, primarily from its Anthropic investments. Net income reached $62.6 billion, or $5.75 per diluted share, versus $18.2 billion a year earlier.

Operating income is the cleaner comparison. That is the point the 24/7 Wall St. column makes, and the filings back it up. Comparing the two EPS figures would mean comparing very different kinds of earnings.

Microsoft's side: monetizing seats it already owns​

Microsoft's strongest evidence for the workflow thesis is Copilot adoption inside products customers already buy.

• Copilot seats. CFO Amy Hood said net paid seat adds more than doubled sequentially, with paid Microsoft 365 Copilot seats now over 30 million.
• Base seat growth. Paid Microsoft 365 Commercial seats grew 6% year over year, mostly in small and medium business and frontline worker offerings.
• Cloud growth. Microsoft 365 Commercial cloud revenue grew 14% as reported, or 16% adjusted for a prior-year period that benefited from in-period revenue recognition.
• ARPU. Hood said premium offerings, including Copilot, E5 and early traction in E7, drove average revenue per user (ARPU) growth.
• Customer mix. For the full year, Microsoft Cloud revenue passed $214 billion. Nearly 90% came from customers outside frontier model companies.
• RPO mix. All sequential commercial RPO growth came from customers outside frontier model companies. RPO grew 25% excluding OpenAI.

These are separate metrics. Copilot paid seats are not total Microsoft 365 seats. Seat counts also say nothing about usage, retention or profitability.

The EY example, with caveats​

In a May 21, 2026 announcement, Microsoft and EY said EY initially deployed Copilot to 150,000 users and recorded a 15% productivity boost. That figure is a Microsoft-published customer claim, not an independent audit. EY is now scaling Copilot through Microsoft 365 E7 to its more than 400,000 people. That describes a rollout in progress, not 400,000 employees already using it.

Microsoft is also an infrastructure company​

Microsoft is not simply "software." Several data points show it is racing to build capacity too.

• Hood said customer demand continues to exceed available capacity.
• Capital expenditures were $41 billion, about two-thirds of it for short-lived assets, primarily CPUs and GPUs.
• Cash paid for property and equipment was $35.8 billion. Operating cash flow was $55.4 billion and free cash flow was $19.6 billion.
• Microsoft Cloud gross margin was 65%, down year over year. Hood attributed that to the mix shift toward Azure, AI infrastructure investment and growing product usage.
• GitHub Copilot usage pressured segment margins, though they improved through the quarter after a June change to usage-based pricing.
• Starting in fiscal 2027, Microsoft is extending the estimated useful life of its datacenters and office buildings from 15 to 25 years. That shifts more future datacenter leases from finance to operating leases. As a result, its calendar 2026 capex expectation adjusts to approximately $175 billion.

The workflow side is not free of infrastructure costs. Every Copilot query lands on someone's silicon.

Amazon's side: capacity, chips and the CPU flywheel​

Amazon's pitch is that AI demand pulls the rest of the cloud along with it. Andy Jassy said growth in AI drives core, because post-training, reinforcement learning and agent tool use are mostly done on CPUs rather than AI accelerators. Amazon's Q2 commentary also says its Graviton CPU offers up to 30–40% better price-performance than other options. That is Amazon's own claim, not an independent benchmark.

Amazon also disclosed scale in its custom silicon and AI businesses. Its AI and chip businesses each surpassed a $25 billion annualized revenue run rate.

The OpenAI link​

OpenAI announced on February 27, 2026 that it would consume approximately 2 gigawatts of Trainium capacity through AWS infrastructure. The commitment spans Trainium3 and next-generation Trainium4 chips. OpenAI's announcement said Trainium4 is expected to begin delivery in 2027. That is a capacity commitment, not proof that the capacity was installed or in use during the quarter.

It does matter for Microsoft readers, because OpenAI is also one of Microsoft's closest AI partners. A multi-cloud OpenAI is a reminder that no hyperscaler holds exclusive claim to a frontier lab's compute.

The cash cost​

Trailing-twelve-month operating cash flow rose 33% to $161.4 billion. But free cash flow fell to a $7.6 billion outflow, driven primarily by a $66.1 billion year-over-year increase in purchases of property and equipment, reflecting AI investment.

Jassy's 2025 shareholder letter explains the logic. He wrote that AWS lays out cash for land, power, buildings, chips and servers typically 6 to 24 months before it can bill for them. He gave useful lives of 30-plus years for datacenters and 5 to 6 years for chips, servers and networking gear. He also said customer commitments underpin the spending. These are management's claims about its own economics, not guarantees.

Where the 24/7 Wall St. framing needs care​

The column's thesis is that Microsoft wins the workflow, Amazon wins raw compute, and the workflow has better economics. Some of its supporting claims are weaker than they look.

• The margin comparison is not like for like. The column sets Microsoft's 41% Intelligent Cloud margin against AWS's 39.4%. Intelligent Cloud is a Microsoft segment containing more than Azure. Both numbers are segment operating margins, and they bundle different things.
• Free cash flow is a company-wide figure. Amazon's negative trailing free cash flow covers the whole company, retail and logistics included. It is not a cloud-only number.
• The AWS backlog figure. The column cites a $496 billion AWS backlog. One secondary investor site repeats that figure. It did not appear in the Amazon filing text I could review, so treat it as unconfirmed.
• The MAI Voice claim. The column says Microsoft's MAI Voice model cut GPU costs in Dynamics 365 by 89%. I could not corroborate this in Microsoft's earnings materials. Without a baseline, it should not be treated as established.
• Capex figures differ by definition. The column's $54.2 billion Q2 capex figure does not match the cash capex number reported in SEC-filed data, which is $53.1 billion. Different capex presentations are easy to conflate.
• Stock and valuation claims. The column's forward P/E and one-year return figures are date-sensitive investment opinion. They are not technology findings, so I am not repeating them here.

What to watch next​

Both companies gave forward-looking guidance.

• Amazon. Q3 net sales guidance is $197 billion to $202 billion, growth of 9% to 12%. Operating income guidance is $22.5 billion to $26.5 billion.
• Microsoft. For Microsoft 365 Commercial cloud, it expects growth of about 16% in constant currency on an adjusted basis, or 15% as reported. Hood said usage-based billing products added alongside per-seat licensing in July should help drive acceleration through the fiscal year.
• Azure. The Wall St. column cites an Azure guide of about 45% for fiscal Q1. I did not verify that figure in the Microsoft materials I reviewed.

What it means for IT admins​

For admins and buyers, the practical point is licensing and cost. Microsoft is layering usage-based billing next to per-seat licensing, as GitHub Copilot's June change shows. The more AI spend becomes consumption-based, the more budgeting and monitoring matter. This is analysis, not a company statement.

On the infrastructure side, capacity constraints at both companies mean new AI workloads may face supply limits. Hood said demand exceeds capacity at Azure, and Jassy said AWS is still capacity constrained.

Bottom line​

"Software versus infrastructure" is a useful lens, but the filings undercut a clean split. Microsoft spent $41 billion on capex in a single quarter and reported margin pressure from AI infrastructure. AWS earned $16.6 billion in quarterly operating income while Amazon's trailing free cash flow went negative. The better question is how each company turns capital into recurring revenue, and how long that conversion takes. The next quarterly reports should show whether Copilot monetization and Trainium adoption are converting as management says.