All capex is not created equal
A framework for what matters
In my 02/05 piece on Alphabet (GOOGL), I noted that it remained one of the most powerful cash-generating businesses on Earth—but its newly disclosed massive spending plan is shifting the narrative from pure growth to questions about capital efficiency and execution risk.
With hyperscaler earnings now complete, it is time to look at this staggering spending landscape — between MSFT, META, GOOGL, and AMZN, the total capital expenditure budget for 2026 is about $640B. Then there is OpenAI, Anthropic, and others in the new LLM Goldrush, who have either raised or seeking to raise tens of billions of capital. For context, India’s budget unveiled on 02/01 was about $630B.
The Dotcom era dearly taught everyone that a firm taking in external capital in itself is not so much of an issue as long as it can generate cash flows to service it — in the foreseeable future, that is. While that holds true, today’s situation is different. The firms embarking on such ambitious spends are not profitless entities hoping to “catch eyeballs” like then, but are cash machines. Their individual capital raises dwarf the aggregate amounts raised then. Nevertheless, there is a striking similarity between them — the overconfidence that they have seen the future and are galloping towards it.
Can these names generate adequate free cash to justify their current valuations? Will their operating environment even allow it? What about the myriad other tech firms who too have raised significant capital relative to their size as the rising tide lifted all boats?
What is my framework to analyze this?
Firstly, it is always about cash — make no mistake. Businesses exist to make profits, responsibly. But profit is an accounting term. Instead, what is more important is how much cash a firm is generating from its operations. It is needed to pay for its capital expenditures (Capex). If some of it is still left after that, it is unencumbered — the free cash flow (FCF). The more positive, the better it is.
Secondly, diversity is strength, here too. Yes, it is that simple. Firms with a multitude of revenue streams sure have a better value proposition than monolines (single-product firms) no matter they are future-defining or merely-following, in technology or any other industry.
Thirdly, regardless of diversification and prowess in technology, innovation, or product development, no firm is immune to the operating environment. Well, you can only take what the market gives you. On my part, I have long held that technology is fundamentally deflationary, because it seeks to do more and better with less. So, there should be no surprises to see the AI juggernaut and the fierce competition it brings along, crush costs and lift productivity. It can hurt the free cash flow generation for all but a few who become so dominant to weather falling unit economics. Hence, the AI capex arms race!
Through that framework, let’s look at the playing field.
I earlier covered Alphabet in some detail (Link) and so let’s move on to the others.
META: At $125B (mid), +66% y/y, the 2026 capex spend targets AI infrastructure, data centers, and “superintelligence” research. At $220B in revenue and ~$165B in expense forecast, FCF could be cut by half or more. Meta has been executing well despite its expensive hobbies, such as Reality Labs (~$19B hit in 2025). Its focus has been almost entirely internal, unlike GOOGL. User engagement is key, therefore. The company has shown its willingness to listen to the market in the past and I wouldn’t be surprised to see that repeat.
MSFT: About $125B of capex has been penciled in, but recent concerns over core software line for lacking frontier AI models has led to MSFT’s re-rating. In line with what I noted earlier, the company enjoys a robust cash flow underpinned by diversified revenue streams. Notably, the strength and stickiness of its Azure cloud cannot be overstated. Besides, a $90B in cash provides adequate cushion.
ORCL: With ~$71B in capex across FY’25-26, the firm went from being a monster cash flow generator to one that may not see positive free cash flow up till 2030. But here is something to consider. While its massive order backlog (~$520B+) may justify the spend, the recent drubbing of software stocks by Anthropic raises uncertainty there. ORCL has been a world leader in SaaS for long. Strategic (circular) investments in AI via OpenAI and NVDA could come handy as offsets but at the cost of higher vulnerability if those fail to scale.
This arms race extends beyond the hyperscalers themselves to the GenAI firms they’re bankrolling—firms whose very survival depends on this circular flow of capital.
Pureplay GenAI: Despite their current appeal to users, leading firms in this space, such as OpenAI and Anthropic, still belong to the speculative category at this stage, given lack of clarity on monetization at scale to justify their massive capital raises. The circular flow of investment dollars — giant tech firms invest tens of billions of dollars in exchange for GenAI firms to commit using their infrastructure and services — makes it all the more head scratching.
But what is clear is that in the case of OpenAI and Anthropic the outcomes are digital — either a world-beating success or face an existential crisis for misallocating capital. The latter case will hurt their capital providers and with their capital entanglements hurt the market and overall sentiment that goes with it.
In the big picture, it is not only about how much capital gets allocated, but also where it lands — because all capex is not created equal.
Disclosure: The author and the portfolios he advises may hold positions in securities mentioned in this article.
Disclaimer: This content is for informational purposes only and does not constitute investment advice. Readers should conduct their own research and consult with financial advisors before making investment decisions.
For portfolio and risk advisory inquiries: gs@macrofireside.com