ShowBiz & Sports Lifestyle

Hot

Will Amazon or Microsoft Solve the AI Energy Bottleneck?

Will Amazon or Microsoft Solve the AI Energy Bottleneck?

Alex SiroisWed, July 22, 2026 at 6:03 PM UTC

0

24/7 Wall St.Quick Read -

Amazon and Microsoft are each committing roughly $200B and $190B in 2026 capex, racing to solve AI's energy bottleneck with opposite strategies.

Microsoft's 46% operating margin, $627B backlog, and P/E of 23 make it the lower-risk AI play while Amazon's free cash flow has collapsed.

Grid interconnection queues will cap Amazon's ability to energize 5 GW promised to Anthropic and 2 GW committed to OpenAI by 2027.

Act now: the analyst who called NVIDIA in 2010 just named his top 10 AI stocks — and Microsoft didn't make the cut. Grab the names FREE today.

Amazon (NASDAQ:AMZN) and Microsoft (NASDAQ:MSFT) both reported quarters shaped by one question: how do you feed AI enough electricity? Amazon guided to roughly $200 billion in 2026 capex, while Microsoft is pacing toward roughly $190 billion this calendar year. Same bottleneck. Very different playbooks.

AWS Hits a 15-Quarter High as Azure Runs Into Its Own Walls

AWS grew 28% to $37.59 billion, the fastest pace in 15 quarters, on an operating margin of 37.7%. Andy Jassy said the chips business is now at a "$20 billion revenue run rate" and would look like "$50 billion" if sold as a standalone. Trainium2 is sold out, and Anthropic locked in up to 5 GW of Trainium capacity.

Microsoft posted Azure growth of 40% and an AI run rate of $37 billion, up 123% year over year. Yet Satya Nadella conceded supply is "expected to remain constrained at least through 2026". Commercial RPO of $627 billion shows demand that Redmond simply cannot serve fast enough.

Building a 2030s Grid vs. Racing to 2028

Amazon added 3.8 gigawatts of power capacity in the past year and is committing tens of billions to custom silicon that Jassy says will "save us tens of billions of dollars of CapEx each year". That is a vertically integrated bet: land, power, chips, racks, all owned.

Act now: the analyst who called NVIDIA in 2010 just named his top 10 AI stocks — and Microsoft didn't make the cut. Grab the names FREE today.

Advertisement

Microsoft is threading a different needle. Its Fairwater data center in Wisconsin came online six weeks early, another gigawatt landed in Q3, and the OpenAI partnership was extended through 2032 with royalty-free IP rights. Redmond leans on NVIDIA plus Maya 200, and squeezes more from what it has: a 40% inference throughput improvement on Copilot's most-used models.

Lens

Amazon

Microsoft

Core Bet

Own the stack, own the power

Efficiency plus OpenAI leverage

Silicon

Trainium, Graviton, Nitro

NVIDIA, Maya 200, Cobalt

Key Vulnerability

FCF collapsed 95% to $1.2B TTM

OpenAI losses hit $3.1B

Power Contracts Will Decide 2027

I will be watching whether Amazon can actually energize the 5 GW promised to Anthropic and the 2 GW committed to OpenAI beginning 2027. Grid interconnection queues will set the ceiling. For Microsoft, the tell is whether Azure growth holds near 39% to 40% as OpenAI absorbs its $250 billion incremental Azure commitment.

Why I Lean Toward Microsoft Until Amazon's Cash Flow Stabilizes

On the data, Microsoft looks like the lower-risk bridge. The stock is down 17.39% year to date while operating margin sits near 46%, and the P/E of 23 looks reasonable given a $627 billion backlog. Amazon's ambition is more thrilling, but with free cash flow at just $1.2 billion TTM and debt climbing to $119.1 billion, I want to see a couple more quarters before the capex flywheel proves itself. Investors weighing higher variance against steadier cash flow will read this trade-off differently; the operational milestones above are what to track next.

Act now: the analyst who called NVIDIA in 2010 just named his top 10 AI stocks — and Microsoft didn't make the cut. Grab the names FREE today.

Contact editorial@247wallst.com for any questions or corrections.

Original Article on Source

Source: “AOL Money”

We do not use cookies and do not collect personal data. Just news.