Is the AI Gold Rush

Already Priced In?

What the latest Big Tech earnings — and the explosive moves in storage, infrastructure, and memory names — tell us about valuations and the question every investor is quietly asking.

STREETWISE ECONOMICS

Applied Economics  ·  Capital Markets  ·  Canada & Beyond

By  ISAAC JONAS    ·    For Streetwise Economics  ·  Social Media Edition    ·    May 8 2026

ON MAY 5, AMD reported quarterly revenue of $10.25 billion — up 38 per cent from a year ago — and the stock jumped 16 per cent the next day. The day after that, Arm Holdings reported its third consecutive year of revenue growth above 20 per cent, with data centre royalties more than doubling. The week before that, Meta raised its 2026 capital expenditure guidance to between $125 billion and $145 billion — almost double its 2025 spend — and the stock fell as much as 10 per cent in after-hours trading. AMD and Arm join NVIDIA, Meta, Micron, and a widening cast of AI-adjacent names that have driven much of the recent market move. Across just under a week, the market made three different decisions about the same underlying story. That tension is the question worth thinking carefully about right now.

I follow United States and Canadian markets closely, and I am not in the business of predicting where stocks go next. Anyone who tells you they know is either selling something or about to be wrong. What I do think is worth doing is stepping back from the daily noise and asking what the recent earnings season actually tells us — about the size of the AI build-out, the discipline behind it, and the prices at which it is being valued today.

THE SCALE OF WHAT IS HAPPENING

The numbers from Q1 2026 earnings are not normal corporate results. They are the financial signature of a generational infrastructure cycle. NVIDIA closed its fiscal 2026 with $215.9 billion in annual revenue, up 65 per cent year over year. Its CEO, Jensen Huang, told analysts in March that he sees more than $1 trillion in purchase orders for Blackwell and the next-generation Vera Rubin platform through 2027. AMD’s data centre revenue grew 57 per cent year over year to $5.8 billion in a single quarter. Arm’s data centre royalty revenue more than doubled. Memory suppliers — including Micron and Western Digital — have seen demand for high-bandwidth memory and enterprise storage lift their results alongside the GPU makers. The recent spin-off of SanDisk has put the storage arm of that supply chain back in front of investors as its own pure-play name.

The infrastructure layer is even broader than the chip names. Sterling Infrastructure, which builds data-centre site work, foundations, and electrical infrastructure across the United States, has seen its share price climb sharply in recent days as investors reprice what it costs to physically build the capacity behind those AI workloads. Similar moves have rippled through power, cooling, and grid-equipment suppliers — companies whose business has nothing to do with software but everything to do with whether the build-out actually happens on time.

On the demand side, the four hyperscalers — Microsoft, Meta, Alphabet, and Amazon — collectively guided to combined 2026 capital expenditure approaching $700 billion. Microsoft alone has guided to roughly $190 billion for the calendar year. These are the largest single-year capital deployments in technology history. They are larger, in nominal terms, than the entire telecom build-out of the late 1990s.

THREE FRAMEWORKS WORTH HOLDING IN MIND

When something this large is happening, three different frameworks for understanding it are worth holding in mind simultaneously. Each of them is partly true and partly limited.

The first framework is the picks-and-shovels view. In every gold rush, the people who reliably make money are not the prospectors but the suppliers — the picks, the shovels, the railroads, the data centres. NVIDIA, AMD, Arm, the data-centre operators, the power and cooling specialists, the cabling and networking companies, the storage suppliers, and increasingly the construction and infrastructure firms that physically build the sites. The argument is that whether or not any individual application of artificial intelligence proves transformative, the infrastructure is being built, and someone has to supply it. This framework has been mostly right for the last two years.

The second framework is the capital expenditure return question. The hyperscalers are spending astonishing amounts of money. Meta has effectively acknowledged that it underestimated the compute capacity it would need, even while ramping aggressively. The question that is now being asked — quietly during the day, loudly during earnings calls — is whether the revenue these investments generate will arrive on a timeline that justifies the spend. Meta’s stock fell on its capex announcement precisely because the market is no longer willing to assume the answer. That signals a healthier market dynamic than blind enthusiasm. It also signals risk.

The third framework is the valuation question. As of early May 2026, the S&P 500 trades at a forward price-to-earnings ratio of approximately 20.9, above the five-year average of 19.9 and well above the ten-year average of 18.9. The trailing P/E sits closer to 27. AI-exposed names trade at higher multiples still — AMD’s stock has more than tripled in the past year, and several infrastructure suppliers have moved 30 to 50 per cent in single sessions on news that would have produced more measured responses two years ago. Valuation does not predict short-term returns, and historically forward P/E has a poor track record as a one-year market timing signal. But valuation does shape the margin for error. At low multiples, you can be wrong about the timing and still do well over the long run. At high multiples, you need the story to play out close to plan.

What matters most is whether the market is correctly distinguishing the companies that will compound from those simply along for the ride — and whether current valuations embed modest assumptions or heroic ones about how quickly data centres and AI infrastructure will transform everything from cloud computing to enterprise software.

THE TWO QUESTIONS I AM SITTING WITH

Two questions sit underneath the surface of every AI investment debate right now, and they are worth being honest about because nobody has a confident answer.

•  Is the demand visibility real or pulled forward? NVIDIA’s $1 trillion order book through 2027 is striking. So is AMD’s commentary about tens of billions of dollars in data centre AI revenue next year. But every cycle of this kind in technology history has produced a moment when demand visibility, which had felt unshakeable, suddenly evaporated. The cleanest signal we have is that customers continue to commit to multi-year purchasing agreements at premium prices. The cleanest counter-signal is that capital expenditure is now growing faster than revenue at the spending end of the chain — and that gap usually closes one of two ways.

•  Is the market correctly separating winners from also-rans? In every infrastructure cycle, the long-term winners are smaller in number than the early enthusiasm suggests. The railroad era produced thousands of companies and a handful of survivors. The early internet era produced a similar pattern. When a single-day move of 30 to 50 per cent is becoming common across infrastructure-adjacent names, it is worth asking whether the market is differentiating carefully or simply assigning a multiple to anything with the word “AI” near it. Whether the AI cycle resolves with a few durable winners and many disappointing also-rans is a question no current valuation can fully price.

DISCIPLINE, NOT PREDICTION

I cannot tell you what the market will do next quarter, and I would not try. What I can offer is the discipline I try to apply to my own thinking. Read the earnings reports rather than the headlines. Pay attention to where capital expenditure is being absorbed and where it is being scrutinised. Ask whether a company’s valuation embeds modest assumptions or heroic ones. Notice that the same data — historic revenue growth, expanding capex, premium valuations — supports different conclusions depending on which framework you bring.

The AI infrastructure build-out is real, and so is the possibility that some of the prices being paid today will be regretted later. Holding both ideas in mind at once is not indecision — it is what serious analysis looks like in the middle of a cycle whose ending nobody yet knows.

DISCLAIMER

This article is provided for educational and informational purposes only. It does not constitute investment advice, a recommendation to buy or sell any security, or an offer of any financial service. Past performance is not indicative of future results. The author may hold positions in some of the companies referenced. Readers should consult a duly licensed financial advisor before making investment decisions. The author and Streetwise Economics accept no liability for actions taken on the basis of this content. All figures referenced are drawn from company earnings releases, FactSet, and reputable financial news outlets as of early May 2026.

ABOUT THE AUTHOR

Isaac Jonas is the founder and Principal Consultant of Streetwise Economics, an applied economics consulting practice based in Abbotsford, British Columbia, Canada. His work focuses on regional economic analysis, labour market intelligence, and capital market commentary, with clients across Canada and Zimbabwe. He holds a Master of Food and Resource Economics and an MA in Resource, Environment and Sustainability from the University of British Columbia, and a BSc Economics from the University of Zimbabwe, where he was a Mastercard Foundation Scholar.

Website: streetwiseeconomics.com    ·    YouTube and Substack: Streetwise Economics

Follow Streetwise Economics on YouTube and Substack for more applied analysis. I am also available on social media — and always happy to discuss these ideas further or to take on commissioned research, applied economic analysis, labour market reports, and policy advisory work for institutional and corporate clients.