Summary
Highlights
IBM released preliminary Q2 results, missing Wall Street expectations with $17.2 billion in revenue (up 1% year-over-year but $660 million below consensus) and adjusted earnings of $2.93 per share ($0.08 below consensus). This led to a 23% stock drop, its steepest in decades, with infrastructure revenue down 7% and consulting broadly flat. CEO Arvind Krishna attributed the miss to clients redirecting capital spending towards AI hardware (servers, storage, memory) at the expense of software and consulting budgets. The market initially perceived this as an IBM-specific execution problem, failing to pivot fast enough for the AI wave.
The video argues that IBM's earnings miss reveals a critical market structure problem: Corporate America is still in the 'infrastructure phase' of the AI cycle. Hyperscalers like Amazon, Microsoft, Alphabet, Meta, and Oracle are projected to spend $700-900 billion on CapEx by 2026, a 36% increase. IBM's clients are also buying hardware for AI, indicating they are in the spending phase, not yet generating returns. This is supported by 'shadow data' showing that 95% of generative AI pilots produce no measurable profits (MIT study) and 42% of enterprises abandon most AI projects (S&P Global), with only 25% of companies achieving expected AI return on investment (IBM's own research).
There's an estimated $600 billion annual revenue gap between what the AI industry spends on infrastructure and what it actually generates in sales, a gap that is widening as CapEx accelerates faster than revenue projections. This pattern is compared to the late 1990s internet infrastructure buildout, where companies like Cisco and Lucent reported record revenues selling hardware, but corporate clients ran out of budget before realizing returns. While AI technology and productivity gains are real, the spending-to-return structure is recognizable.
Hyperscalers selling infrastructure are thriving, with AWS growing 28%, Google Cloud up 63%, and Microsoft's AI business crossing $37 billion annually (up 123%). These companies are 'selling the shovels in the gold rush.' However, IBM, which sells 'guidance on where to dig' (consulting and software), is struggling because corporations are still buying shovels, not guidance or productivity. This highlights the distinction between companies profiting from the infrastructure phase and those dependent on the returns phase.
The S&P 500 earnings forecasts are up 24% year-over-year due largely to semiconductors and AI infrastructure, yet the index has been flat for two months. This suggests the market is questioning the reality of the AI earnings story beyond just infrastructure. IBM's earnings partially answer this: buyers are still acquiring hardware, not productivity. The gap between the infrastructure phase and returns phase poses a significant risk to technology earnings models assuming AI-driven margin expansion by 2027. Investors should evaluate whether their tech holdings sell into the infrastructure or returns phase.
The full earnings call on July 22nd will be crucial. If IBM's CEO attributes the shortfall to a timing issue, the damage might be contained. However, if guidance shows enterprises structurally redirecting away from consulting towards hardware long-term, it signals a deeper repricing for legacy tech companies relying on AI consulting. A critical concern is what AI budgets will look like in 2027 if the 95% of pilots still haven't produced measurable profits, leading to tough conversations about ROI and further impact on consulting services like IBM's.