The technology trade has arrived at a genuine crossroads, and the contradictions inside it are growing harder to ignore. On one side sits a software sector in a tailspin, where legacy names are struggling to prove they have not been disrupted by the very artificial intelligence they helped create. On the other side stand the hyperscalers, preparing to detonate a roughly $700 billion capex bomb on the market — assuming, that is, they can actually get their data centers built through a thicket of community protests, permitting hurdles, and power shortages. Few earnings weeks in recent memory carry as much weight in resolving that tension.
The Hyperscaler Showdown
Wednesday evening will produce results from four of the most consequential names in the cloud: Amazon, Alphabet, Meta, and Microsoft. Of these, Microsoft has so far appeared to struggle the most with resources, particularly in obtaining enough chips to run its cloud services. Alphabet, by contrast, has generally led the pack, and there is little reason to expect that pole position to change. Microsoft will likely continue to grapple with capacity constraints while Alphabet maintains its lead.
The headline figure for the year — somewhere between $680 billion and $720 billion in expected capex across these companies — is staggering enough to invite a serious question: at what point does aggressive growth start to look like irrational overspending? Recent reporting that OpenAI may be missing some of its internal targets serves as a quiet foreshadowing. If and when OpenAI eventually files to go public, its CFO will have to deliver quarterly answers to public investors about growth and the eventual path to positive free cash flow. The company has already pledged to spend in excess of $852 billion, an amount that today exists more as ambition than as audited reality.
Should this year bring IPOs from OpenAI, Anthropic, and SpaceX — the latter representing Elon Musk's combined corporate vehicle — the requirement to answer to public markets every ninety days will function as a profound reality check. SEC filings have a way of forcing concreteness onto numbers that today float somewhere between $700 billion and a trillion dollars depending on who is doing the projecting. The rubber meets the road only when these companies must justify their spending to public investors.
Roadblocks on the Ground
Even setting aside financial discipline, the physical buildout faces real friction. Project Blue in Arizona has already drawn organized, in-person protest, with residents lining up to oppose the energy and water demands placed on their communities. In a midterm election year, that local discontent is far from trivial. Politicians listen to constituents in their counties, and the disfavor expressed at town halls and county meetings translates directly into political pressure on permitting and approvals.
Beyond the politics, the grid itself is a constraint. Companies need permits to connect, and once permitted, sufficient grid capacity must actually exist to deliver the power. The deployment timeframes laid out by Microsoft, Alphabet, and others — already vague and dependent on a great deal of hope — extend out to 2028 and beyond. Local opposition to energy and water consumption in residents' backyards represents a genuine, material threat to those schedules.
The Other Half of the Equation: Software in the Rut
If Wednesday belongs to the cloud giants, Thursday belongs to the software vendors. Reddit, Roblox, Atlassian, and Twilio will all report after the close, and investors will be scanning each release for one specific signal: evidence of incremental revenue from AI. Anything qualitative or quantitative that demonstrates these companies are actually monetizing AI — rather than being eaten by it — would help calm the pervasive fear that traditional software is becoming irrelevant in an age dominated by Anthropic and OpenAI.
Atlassian provides a striking illustration of the disconnect. The company is trading at dot-com era lows even as it grew cloud revenue by 26%. That tension raises an unavoidable question: is this a valuation problem, or is the market fundamentally afraid that agentic AI will replace the seats and licenses these companies have built their businesses around?
The honest answer is that both are at work. Part of the malaise is the existential threat AI poses to seat-based software economics. The other part is what might be called bad karma from 2020 and 2021, when names like Atlassian traded at extraordinary valuations on the back of multi-year forward projections. Those projections have evaporated, and no one is quite sure what model should replace them. The result is a sector trading at nominally very low multiples without any consensus on whether that cheapness is opportunity or warning. Last week's plunge in ServiceNow shares, on a relatively good earnings report and with a much cheaper valuation than its historical norms, drove the point home. The stocks are cheap because investors cannot confidently answer two questions: will these companies survive the AI onslaught, and how are we supposed to model their free cash flow years into the future when the 2021 playbook no longer applies?
The Bellwether to Watch
If a single name is going to define the health of what might be called the AI 2.0 trade in this earnings cycle, the most interesting candidate is Twilio. The company reports Thursday alongside the other three software names, and its CEO, Khozema Shipchandler, has spent more than a year refashioning both the growth profile and the profitability of the business. A series of analyst upgrades has rolled in over recent weeks, and the stock has rallied meaningfully over the past thirty days.
There is a real possibility that Twilio delivers a print demonstrating exceptional growth and profitability inside an otherwise dismal software trade. If it does, the recent momentum has a path to sustaining itself, and the broader software cohort may finally have a credible counter-narrative to the AI displacement thesis.
Conclusion
The week ahead is less about any single number and more about whether two opposing tensions can resolve. Hyperscaler capex is on the verge of crossing thresholds that demand a much more concrete accounting of returns, particularly as private AI giants approach public markets and as community resistance and grid constraints begin to bite. Meanwhile, the software vendors most threatened by this AI buildout must demonstrate that they are participating in the upside rather than being consumed by it. Whichever way the prints fall, the era of fantasy numbers and unbounded forward projections is drawing to a close, and a more disciplined chapter of the AI trade is beginning to take shape.