Morgan Stanley forecasts $805B AI capex
- Morgan Stanley raised its forecast for 2026 hyperscaler capital spending to $805 billion across Amazon, Alphabet, Meta, Microsoft, and Oracle, reigniting the AI buildout story. - The punchline is speed: the bank also sees $1.1 trillion in 2027, while David Sacks framed AI capex as a 2.5%–3% GDP tailwind. - That matters because recent earnings already showed capex jumping fast, suggesting the spending race is moving from forecast to operating reality.
AI capex is turning into the main plot of the tech economy. Not app launches. Not chatbot demos. Actual concrete, steel, power contracts, GPUs, and data-center leases. What changed this week is that Morgan Stanley pushed its forecast for hyperscaler capital spending to $805 billion in 2026 for Amazon, Alphabet, Meta, Microsoft, and Oracle — and to $1.1 trillion in 2027. (cryptobriefing.com) ### What is this number actually measuring? This is capital expenditure — the money these companies spend to build the physical base layer of AI. Think servers, networking gear, land, buildings, cooling systems, custom chips, and the electrical infrastructure to run them. It is not just “AI spending” in the vague sense. It is the industrial buildout underneath the software story. (morganstanley.com) ### Why are these five companies the center of it? Because they are the firms with the cash flow, cloud businesses, and customer demand to build AI capacity at absurd scale. Amazon and Microsoft already sit at the center of enterprise cloud. Alphabet is pushing hard through Google Cloud and its own models. Meta is spending to train and deploy AI across ads, feeds, assis(morganstanley.com)AI customers need more compute than the traditional cloud leaders alone can supply. (cryptobriefing.com) ### Why did the forecast jump now? Because company guidance keeps moving up, not down. Alphabet just raised its 2026 capex target to as much as $190 billion. Meta lifted its 2026 capex range to $125 billion to $145 billion. Amazon’s first-quarter capex jumped to roughly $44 billion, driven mainly by AWS and generative AI infrastructure. Those are not “maybe someday” numbers — they are current spending signals that force analysts to re-mark the whole curve higher. (cnbc.com) ### Why does Wall Street care so much? Because this is the cleanest evidence that AI demand is not just consumer hype. If the biggest cloud companies are willing to spend hundreds of billions before the revenue fully shows up, they are telling you they see durable demand from enterprises, model builders, and developers. Basically, capex is becoming the proof point. The catch is that invest(cnbc.com)y one of the largest infrastructure cycles the industry has ever seen. (seekingalpha.com) ### Where does the GDP angle come from? David Sacks seized on the forecast because capex flows straight into real economic activity. Data centers need construction crews, electrical equipment, transformers, fiber, backup power, semiconductors, and a lot of engineering labor. He argued AI capex could add about 2.5% to GDP growth this(seekingalpha.com) an investment boom. (benzinga.com) ### What are the risks hiding underneath? Power is the big one. You can order GPUs faster than you can build substations. Margins are another. These companies can fund the buildout, but free cash flow gets squeezed when capex ramps this hard. And then there is execution risk — if model demand, pricing, or enterprise adoption disappoints, some of this capacity could look premature. (msn.com) ### So what is the real takeaway? The important shift is that AI has moved from a product cycle to an infrastructure cycle. Morgan Stanley’s $805 billion forecast matters less as a precise number than as a signal: the biggest tech firms are acting like AI will require utility-scale buildouts for years, not quarters. (cryptobriefing.com)
