The last earnings season delivered a verdict. Not on AI itself. On who is allowed to burn capital for it.
Alphabet guided $205 billion in capital expenditure. Its free cash flow went negative for the first time since 2004. Market reaction: 6% drop. Tesla spent 142% more on capex year-over-year. Earnings missed. Stock cratered. ServiceNow spent nothing extraordinary on infrastructure. Raised subscription revenue 24.5%. Market shrugged off a 3.7% post-earnings dip and kept buying. Intel reported its fastest revenue growth in 15 years on the back of Gaudi AI chips. It got labeled a 'winner.'
Context is simple. We are in a bull market for AI hype. But the market is no longer buying promises. It is auditing receipts. The ledger keeps score. And the ledger is showing that some of the biggest names in tech are burning capital faster than they can convert it into profit. This is the first real test of capital efficiency in the AI arms race. The results are not pretty.
Core: The Mechanical Cruelty of Capital Expenditure
Let's dissect the numbers coldly. Alphabet's cloud revenue grew 82% year-over-year. Impressive. No one disputes that. But the capital expenditure required to generate that growth is staggering. The $205 billion capex guidance is for the year. That is roughly three times the incremental cloud revenue they are adding. In other words, for every dollar of new cloud revenue, Alphabet is spending three dollars on infrastructure. That is not a business. That is a capital sinkhole with a growth story attached.
Code is truth. Intent is fiction. The balance sheet is the on-chain data of corporate behavior. Alphabet's free cash flow flipped negative. That means they are now borrowing or diluting to fund the build-out. They are minting future equity to buy GPUs today. The market smelled the dilution risk before the numbers were printed. The stock sold off.
Tesla is worse. Capital expenditure growth of 142% to deliver 7.2% revenue growth. That is negative unit economics masked by a rising tide of vehicle deliveries. The core automotive margin is being squeezed by price cuts. The FSD software revenue that was supposed to be the high-margin savior? Still not materializing at scale. EPS of $0.33 missed expectations. The market punished Tesla hard. The mechanical truth is that Tesla is spending like a growth company but earning like a cyclical manufacturer. That gap will not be filled by hype alone.
ServiceNow took the opposite approach. Their AI assistant, Now AI, is integrated into an existing SaaS platform. No massive data center build-out. No proprietary silicon. They leverage third-party models and fine-tune. The result: subscription revenue grew 24.5%, cRPO (remaining performance obligations) grew 21% to $13.2 billion, and they are free cash flow positive. The capital expenditure is negligible relative to revenue. ServiceNow is the poster child for 'capital-light AI monetization.' The market recognized that. Their stock held up.
Intel is a different case. Revenue up 25% in a quarter. Driven by Gaudi AI accelerators. Intel is not the leader—Nvidia is. But Intel is the cost-effective alternative. The 'open system' narrative is resonating with hyperscalers who want to reduce dependence on Nvidia's CUDA lock-in. Intel's turnaround is real, but fragile. The capital expenditure is already in the ground from previous years. Now they are harvesting. The market rewarded that because it smells a multi-year replacement cycle.
Where the Numbers Hide the Truth
The published numbers don't tell the whole story. Alphabet's cloud revenue growth is real, but how much of it is from large enterprises migrating mission-critical workloads versus AI startups burning venture capital on compute? The latter is fickle. When the venture capital tap turns off, those workloads disappear. Alphabet is building infrastructure for a demand wave that may be 90% startup-driven. The actual enterprise adoption of AI is still in pilot phase. Alphabet is betting the farm on pilots converting to long-term contracts. If that conversion rate drops below expectations, their $205 billion capex becomes stranded.
Tesla's capex is not just for AI. It includes factory expansions for Cybertruck, semi-truck, and future models. But the market is pricing it all as an AI bet. That is dangerous. If the AI returns—full self-driving—are delayed further, the capex will have to be written off against automotive margins. The risk is not just underperformance; it is a permanent capital impairment.
ServiceNow's cRPO ratio of 3.4x (cRPO to subscription revenue) indicates that future revenues are already locked in. That is a massive moat. Their customers cannot easily leave because the platform is deeply embedded in IT workflows. AI is just a feature that increases stickiness. ServiceNow didn't need to build a new capital-intensive moat. They just deepened the existing one.
Contrarian: What the Bulls Got Right
The bulls were not wrong about the opportunity. AI is a generational shift. But they were wrong about the timeline and the capital intensity. For Alphabet, the argument is that cloud infrastructure has long payback periods and the investments will eventually yield high returns. That may be true. But the market is punishing the uncertainty. The bull case hinges on a multi-year horizon that investors are no longer willing to discount at high rates. The contrarian truth is that Alphabet's cloud business could indeed become a cash cow in 2027. But by then, the capital markets may have already re-rated the stock downward multiple times.
For Tesla, the bull case is that Full Self-Driving will unlock billions in high-margin software revenue. But the evidence is not there yet. Subscription rates are low. Regulatory approval is uncertain. The bulls got right that Tesla is a leader in EV manufacturing at scale. They got wrong that AI would be accretive within two years.
ServiceNow bulls were right about capital efficiency. They were also right about AI being an upsell catalyst, not a cost center. The contrarian blind spot is competition: Microsoft, Salesforce, and others are also embedding AI into their platforms. ServiceNow's NRR could compress if competitors offer AI features at lower prices. But for now, the data supports the bull case.
Intel bulls were right that the AI chip market is big enough for multiple players. Nvidia cannot serve all demand. The blind spot: Intel's Gaudi is still a generation behind. If Nvidia cuts prices aggressively, Intel's window of opportunity may close. The market is pricing a successful turnaround, but execution risk remains high.
Takeaway: The First Big Test Has a Scorecard
The earnings season just became a referendum on capital efficiency in the AI arms race. ServiceNow and Intel passed. Alphabet and Tesla failed. The lesson is mechanical: capital deployed into infrastructure must produce proportional revenue within a reasonable timeframe. If it doesn't, the market will reprice the stock downward, regardless of the long-term narrative.
This is the same principle I applied when auditing token contracts in 2017. Beautiful code that hides a reentrancy vulnerability is still broken. Beautiful growth numbers that hide negative free cash flow are still broken. The ledger keeps score. And the ledger is saying: the bull market for AI hype is over. The bear market for capital efficiency has begun.
Gas fees don't lie. Neither do capital expenditures.
