Author: Ada, Shenchao TechFlow
On April 23, Texas Instruments' stock price achieved its best single-day performance since 2000. On the same day, ServiceNow recorded its largest single-day drop in history.
The same earnings season, the same trading day, two signals in opposite directions. The market is drawing a line with real money—below this dividing line, AI infrastructure wins, and AI upper-layer applications lose.
Chipmakers Are Laughing, Subscription Sellers Are Crying
On Thursday, Texas Instruments delivered a nearly flawless report card. Q1 revenue was $4.83 billion, up 19% year-over-year, with EPS of $1.68, far exceeding market expectations of $1.40. Data center revenue surged 90% year-over-year. Industrial and analog chip businesses recovered across the board.
The stock price rose 18% that day. Bank of America directly upgraded its rating from neutral to Buy, raising the target price from $235 to $320.
Its performance expectation for the second quarter is revenue ranging from $5 billion to $5.4 billion, with the midpoint exceeding Wall Street expectations by over 10%. Management stated that the recovery in industrial and data center sectors is "still accelerating."
After hours, Intel dropped another bombshell. Q1 revenue was $13.58 billion, far exceeding expectations of $12.42 billion. Non-GAAP EPS was $0.29, while market expectations were $0.01—29 times higher than expected. Data center business revenue grew 22%, reaching $5.1 billion.
Intel's stock price surged over 20% after hours, breaking through the historical high set during the dot-com bubble in 2000. After the U.S. government took a 10% stake last year, the stock has risen over 80% this year.
The狂欢 (celebration) of the chip industry has an underlying logic. Because AI is not air—it consumes electricity, requires chips, and occupies data centers. From Nvidia's GPUs to Texas Instruments' analog chips, to Intel's CPUs and advanced packaging, the entire AI infrastructure food chain is being called to "get on board."
The semiconductor ETF (SMH) has risen nearly 28% this year, with a 22% gain in April alone. The S&P 500 rose 4% during the same period.
On the other side of the coin, the software sector is experiencing a massacre.
ServiceNow plummeted 18% that day, its worst single-day performance in history. IBM fell nearly 10%. Then it spread rapidly: Salesforce, Workday, Oracle, Adobe, Palantir—all declined sharply. The iShares Expanded Tech-Software ETF (IGV) fell nearly 5% that day.
The most ironic part is that IBM and ServiceNow's earnings reports were not bad. IBM's revenue exceeded expectations, and ServiceNow also beat estimates. But the market doesn't care. The market is pricing in a deeper fear: your moats are being eroded by AI.
SaaS Apocalypse
This is not a one-day event.
Over the past few months, a new phrase has been circulating in tech, venture capital, and public markets: "SaaSpocalypse," the SaaS apocalypse. Since February, the software stock crash hasn't stopped. So far, approximately $2 trillion in enterprise software market value has evaporated.
Salesforce is down over 30% this year. Workday is down 33%. Adobe is down 27%. Even Microsoft has fallen 16%. The software ETF (IGV) has dropped from a historical high of $117 to around $82, entering a technical bear market. The forward P/E of the software sector has fallen below the overall level of the S&P 500 for the first time since the mid-2010s.
Why?
The core logic is just one sentence: AI enables companies to do it themselves.
The traditional SaaS business model charges per seat: if you have 100 employees using my software, you buy 100 licenses. But with the arrival of AI Agents, one Agent can replace the work of 10 employees. The number of seats decreases, and so does the subscription fee.
Even more critically, some companies are starting to use AI directly to build internal tools, bypassing the SaaS middle layer. Previously, buying a CRM suite might cost $300 per person per month; now, having AI write an internal system could cost only a tenth of that.
In other words, the moat for SaaS used to be switching costs and user stickiness. But AI short-circuits both at once. Switching costs are lower because AI can automatically handle data migration; stickiness is lower because users no longer need to learn new tools.
Narrative Shift
Look at two sets of data.
Year-to-date, the semiconductor index has risen about 40%. The software index has fallen over 13%. The gap between the two exceeds 50 percentage points.
What does this mean? Capital has not left the tech industry. It has simply made a precise rotation within the tech industry: from the application layer to the infrastructure layer.
The logic now is simple. If I want to bet on AI, I buy chips, because no matter which AI wins, they will need chips. But I might not necessarily buy SaaS.
This is the market's cruelty. Chips are a确定性赌注 (certain bet)—regardless of how AI develops, the demand for computing power only increases. But software is a conditional bet: it only has value if AI cannot completely replace software and if software companies can successfully transform.
Both premises are uncertain, and capital hates uncertainty.
However, attributing all of ServiceNow and IBM's declines to AI threats is not entirely fair.
ServiceNow's CFO Gina Mastantuono mentioned a very specific reason during the earnings call: order delays due to conflicts in the Middle East. New orders from clients in the Iran direction were postponed, which directly dragged down subscription revenue for the quarter.
IBM's issues also have specific explanations. Software business growth slowed from 14% last quarter to 11.3%, mainly due to drag from Red Hat's cloud business. Overall revenue growth slowed from 12.2% to 9%. Subsequently, IBM maintained its full-year guidance unchanged without an upward revision.
But the market doesn't care about these details at all.
In an environment where everyone is worried that "software is dying," any less-than-perfect earnings report will be interpreted as "See, it's starting." Once this sentiment takes hold, data becomes unimportant. The narrative is what matters.
And the current narrative is: AI is the hunter at the top of the food chain, and SaaS is the prey at the bottom.
Behind the Surge
Behind Texas Instruments' 18% surge lies a not-so-pretty number: a P/E ratio exceeding 50 times. Over the past three months, insiders sold $26.5 million worth of stock and bought none.
Intel's forward P/E is 120 times, more than four times that of the S&P 500. One valuation firm给出的内在价值 (gave an intrinsic value) of $27, while the stock is trading around $67—a 147% premium.
It's clear that current chip stock prices have already discounted growth for the next three years. Buyers are not buying this quarter's performance; they are buying faith that AI capital expenditure won't stop.
This year, the combined AI capital expenditure of the four tech giants exceeds $500 billion. Google alone is burning $180 billion. As long as this capital expenditure cycle doesn't stop, chip stocks have support. But what if the giants suddenly find the return on this spending isn't high enough?
Don't forget, when Alphabet last announced $180 billion in capital expenditure, its stock price plummeted 6% after hours. The market's reaction was: "We know you're building AI, but we're starting to worry if you can hit the ball."
Zooming out, the earnings divergence on April 23 reveals a larger structural shift.
AI value capture is migrating downward. From software-layer subscription fees, it渗透 (penetrates) down to hardware-layer chip fees, energy fees, and data center fees. The profit distribution map of the entire tech industry is being redrawn.
When chipmaker stock prices hit records since 2000 and subscription sellers fall to yearly lows, the market is really only saying one thing: I know AI is real, so I'm buying infrastructure. But is AI useful? I'm not sure yet, so I'm not buying applications.
How long will this split last? No one knows. But the next earnings season can serve as a reference. If the giants' AI capital expenditure continues to increase, chip cash flow can continue to support valuations. But if a giant suddenly hits the brakes, this dividing line will reverse.
Until then, the chip celebration continues, and the SaaS funeral continues.
