Dear readers,
On Tuesday we wrote that the market was rewarding proven earnings power over product roadmaps — that Schwab’s deposit-driven cash machine was crushing fintechs still pitching optionality. The PCE print we flagged as the week’s defining data point landed hot, at 3.8 percent, the strongest reading in three years. And now the same ruthless sorting mechanism has arrived in enterprise software, where Salesforce just delivered the quarter that crystallizes the central contradiction of AI investing in 2026: the numbers can be excellent and the stock can still get punished.
Salesforce: Record Revenue, Falling Stock
Salesforce posted first-quarter fiscal 2027 revenue of $11.13 billion, up 13 percent, with adjusted earnings of $3.88 per share. Both figures cleared Wall Street estimates by a comfortable margin. The stock dropped in after-hours trading anyway, extending its year-to-date decline past 30 percent. The trigger was narrow: second-quarter revenue guidance of $11.27 to $11.35 billion came in just below the $11.36 billion consensus. In a different market, that hairline miss gets shrugged off. In this one — with ten-year Treasuries parked at 4.48 percent and the Fed locked in place — it is enough to send sellers running.
CEO Marc Benioff responded with a $25 billion accelerated share repurchase program. The buyback is enormous, but it reads more as defensive reassurance than offensive confidence. When management’s loudest move is financial engineering rather than operational guidance, the market notices.
The Agentforce Paradox
The deeper anxiety has nothing to do with one quarter’s guidance. It is about whether autonomous AI agents — systems like those built by Anthropic — will erode the per-seat licensing model that has powered enterprise software for two decades. Salesforce is not standing still. Its Agentforce platform has reached $1.2 billion in annual recurring revenue, growing at 205 percent. Those are impressive numbers in isolation. They also represent less than five percent of total revenue. The old business is a supertanker; the new business is a speedboat. The question Wall Street cannot answer yet is whether the speedboat scales fast enough before the supertanker’s course becomes obsolete.
Making that transition profitable is another problem entirely. Data from Entelligence estimates that only 18 cents of every dollar invested in AI currently flows into production value. The rest disappears into debugging and rework. For a company trying to migrate millions of enterprise customers from traditional subscriptions to agent-based workflows, that efficiency gap is not an abstraction. It is a margin headwind with a dollar sign attached.
The Cost Problem Gets an Engineering Answer
Salesforce AI Research is attacking the cost side directly. Its new xRouter system uses reinforcement learning to optimize how large language models are orchestrated — routing queries to the cheapest capable model rather than defaulting to the most powerful one. The company projects query cost reductions of 15 to 30 percent and response times cut in half. It is a technical solution to what is fundamentally an economic problem: AI applications cannot achieve software-like margins if every customer interaction burns through expensive compute. The companies that solve token economics will define the next generation of enterprise software winners. The ones that do not will find their AI features are impressive demos attached to deteriorating unit economics.
Geopolitics and Inflation Tighten the Vise
The macro backdrop is offering no relief. Escalating U.S.-Iran tensions have kept Brent crude elevated around $96 per barrel, and those energy costs are feeding directly into the inflation data. The April PCE reading of 3.8 percent — the Fed’s preferred gauge — removes any near-term prospect of rate cuts. For Fed Chair Kevin Warsh, the number validates caution. For equity markets, it means the S&P 500 faces stiff resistance around 5,600, weighed down by a ten-year yield that refuses to budge from 4.48 percent.
High-risk assets are absorbing the worst of it. Following renewed U.S. airstrikes on Iranian targets, the crypto market saw nearly $1 billion in liquidations. Bitcoin briefly broke below $73,000, and spot ETFs recorded single-day outflows exceeding $733 million, led by BlackRock’s IBIT fund. Meanwhile, the trade damage is spreading: Eurostat data shows EU exports to the United States collapsed more than 30 percent in the first quarter.
What This Means for Portfolios
The AI investment thesis is entering its most demanding phase. Buying Nvidia at any price because data centers needed GPUs was straightforward. Figuring out which software companies will actually convert AI capabilities into durable, profitable revenue streams is a different exercise entirely. Salesforce’s quarter illustrates the bind: strong enough to prove AI demand is real, not strong enough to prove the business model transition will be painless.
For investors, the framework is shifting from “who is spending on AI” to “who is earning from AI” — and earning at margins that justify current valuations. In an environment where 3.8 percent inflation keeps rates elevated and geopolitical risk premiums refuse to compress, that filter will only get stricter. The infrastructure trade was the easy part. The software trade is where conviction gets tested.
Best regards,
The StocksToday.com Editorial
