The SaaSpocalypse Is a Multiple, Not a Story: Why Enterprise Software at 2x Sales May Be the Trade That Works in Reverse
Enterprise software has been cut in half on revenue that hasn't collapsed. The thesis: this is multiple compression, not secular decay — and the $3 trillion AI capex bill will eventually force investors to demand ROI from the incumbents.
On 4 June 2026, Salesforce CEO Marc Benioff sat across from the allin hosts and delivered a line that captured an entire quarter of capital markets: "Not my first SaaSpocalypse." His company had shed roughly 37% of its value, the panel noted — a $90 billion erosion in market capitalisation that came not because Salesforce's fundamentals cracked, but because the entire enterprise software complex was rerated downward in lockstep. Workday was down 45%, with $180 billion in market cap erased. ServiceNow, off 42%. The pattern held across the top ten enterprise software names. Most of them were now trading at roughly two times annual sales — a multiple the sector had not seen in a generation, despite what several chief executives described as strong underlying quarters.
This is not a revenue story. It is a multiple-compression story, and the distinction matters enormously for how the trade resolves.
The compression is uniform, the fundamentals are not
The argument for treating the SaaS sell-off as mechanical rather than fundamental rests on a simple observation: companies with dramatically different business models, customer concentration, and growth profiles are being punished with near-identical severity. If the market were pricing in deteriorating unit economics, the dispersion would be wide. It is not. The compression is broad-based because the input that changed is the discount rate investors are willing to apply to long-duration software cash flows — not the cash flows themselves.
Salesforce is on track for $46 billion or more in revenue this year, generating more than $16 billion in annual cash flow on a workforce of 83,000. The company has authorised a $50 billion buyback — among the largest in corporate history. Workday, ServiceNow, and the rest of the cohort can point to similar operating health. Yet the multiple has compressed anyway, because the question on the buy side is no longer "how durable is this revenue" but "what is the cost of capital when AI capex is competing for the same dollar?"
The $3 trillion capex bill is the actual variable
Here is the frame that resolves the apparent contradiction: the AI buildout has consumed an enormous share of incremental investment dollars, and that capital has to earn a return somewhere. If the hyperscalers and frontier labs spend on the order of $3 trillion over the coming years on chips, data centres, and inference infrastructure, then the returns have to come from somewhere downstream. The most obvious downstream beneficiaries are the software platforms that sit between the model layer and the enterprise buyer.
The current trade — selling SaaS because AI will eat it — implicitly assumes that the incumbents capture none of that downstream value. That is a bet, not a fact. Salesforce is already deploying AI agents at scale: the company reported calling back 50,000 people in a single week from a pool of 20 to 30 million never-called-back leads. ServiceNow is integrating agentic workflows into its platform. The companies that own the customer relationship, the data, and the integration surface are not obviously the losers in a world of cheaper inference. They may, in fact, be the only credible distribution channel for AI to reach the enterprise.
The real bifurcation: low end versus high end
The split inside the SaaS complex is now stark. The low end of the market — single-feature point solutions, undifferentiated workflow tools, software that a large language model can replicate with a weekend of prompt engineering — is finished. There is no safe space at the bottom. The high end, where the large monoliths operate with deep customer integration, regulated data residency, and thousands of bespoke configurations, is comparatively safe. The enterprise deployment problem turns out to be far harder than building the model.
OpenAI's own $4 billion investment in a deployment-company joint venture, structured to deliver a 17.5% preferred guaranteed return, is effectively an admission that the model layer cannot reach the enterprise without a partner. The deal stands up a direct competitor to Accenture, Deloitte, EY, PwC, and Cognizant — the systems integrators who have historically been the bridge between software vendors and the Fortune 500. That admission is bullish for incumbents who already own the customer relationship. It is bearish for the long tail of SaaS companies that depended on the integrator channel to reach buyers they could never touch directly.
The China question is downstream, not central
A separate but related thread from the same discussion: the question of whether the United States should sell Nvidia's latest chips to China is increasingly academic. Benioff's argument, delivered bluntly, is that the export controls are "ego gratification" rather than strategic — Chinese models are now as competitive as US models, having learned to make those models without the highest-end chips. The structural reality is that the two countries are negotiating a bipolar carve-up: rare earths, energy, oil, and spheres of influence in Latin America and the Middle East. China has moved roughly 500 million people out of poverty and into the middle class; that domestic market alone sustains a formidable AI industry regardless of chip access.
The bipolar framing matters for SaaS because it tells investors where the durable demand pools sit. If the US concedes influence in parts of Asia and gains leverage over energy and rare earth supply chains, the enterprise software map follows the new geography. Salesforce's exclusive partnership with Alibaba — a one-of-a-kind arrangement in which all data must remain in China and be accessible to Chinese authorities — is a preview of how global software companies will have to structure themselves to serve both blocs.
Forward view: the trade that works in reverse
The asymmetry of the current setup is what makes the contrarian case plausible. If the AI capex cycle disappoints — if the $3 trillion does not produce proportional revenue downstream — the capital will rotate back into cash-flowing software incumbents trading at a fraction of their historical multiples. If the AI capex cycle succeeds, those same incumbents are the most likely distribution channel for the productivity gains. Either way, the bet against enterprise software at two times sales is a bet against the productivity story the AI trade is premised on.
The risks are real. A super El Niño cycle, with oceans releasing the 11 million terawatt-hours of excess heat energy currently stored below the surface — equivalent to 500 years of total human energy consumption at today's 25,000 TWh annual run rate — could disrupt the macro backdrop badly enough to make multiple expansion a secondary concern. A food crisis in India, where 150 million farmers depend on monsoon output feeding 1.5 billion people, compounded by a fertilizer shortage from any Iran-Hormuz disruption, would force capital into defensive positioning regardless of the software thesis.
But the base case is that the SaaS sell-off of mid-2026 is a multiple event that the market will eventually walk back. The trade that works in reverse is the one that pays investors for waiting while the AI capex cycle either fails on its own terms or succeeds through the incumbents. Enterprise software at two times sales is the asymmetric position. It requires patience. It does not require the revenue to actually collapse.
Wire provenance
This editorial synthesis draws on the following public wire/social posts:
- https://www.youtube.com/watch?v=jJRAvZNGUvI