Alphabet’s decision to raise AI-related capital spending guidance again points to one thing, according to WhiteLine Daily: compute supply remains the core bottleneck. The note added that if reports about Google sourcing more than 3 million TPUs from Intel eventually materialize, Intel could become the largest marginal surprise beneficiary in this expansion cycle.
CapEx guidance moves higher again
Alphabet reported second-quarter capital expenditures of $44.9 billion, nearly double from a year earlier. Combined with $35.7 billion in the first quarter, first-half spending reached about $80.6 billion. The company also lifted its 2026 CapEx guidance from $180 billion-$190 billion to $195 billion-$205 billion.
Using the $200 billion midpoint, Alphabet would need to spend close to $120 billion in the second half, or about $60 billion per quarter. Management’s explanation was not that demand had cooled. Instead, the company said capacity had expanded significantly over the past three years, while demand still exceeded supply. It also reiterated that capital expenditures would rise meaningfully in 2027.
Industry supply-chain estimates cited in the note push the scenario further. If roughly 7GW of TPU capacity and 1.5GW of GPU capacity come online in 2027, Google’s full-stack spending could reach $300 billion-$350 billion. WhiteLine Daily stressed that this is not Alphabet’s formal guidance. It is a high-case estimate reverse-engineered from chip volumes, racks, power usage and data-center costs.
The point of that exercise, the report said, is not to claim Google will definitely spend $350 billion. It is to show that concerns over CapEx peaking immediately may still be premature.
Spending is starting to meet revenue
WhiteLine Daily said Google is willing to keep spending because demand and revenue are beginning to catch up. Google Cloud posted $24.8 billion in revenue in the second quarter, up 82% year over year and well above market expectations. Backlog reached $514 billion. Gemini model APIs processed about 22 billion tokens per minute, up from 16 billion in the prior quarter.
The company also confirmed for the first time that it has begun directly recognizing TPU sales revenue. In WhiteLine Daily’s framework, Google is building a three-layer stack:
- First, it sells compute through TPUs and data centers.
- Second, it sells models and enterprise services through Gemini APIs, Workspace and Cloud.
- Third, it integrates AI into Search, Android and YouTube to improve advertising and traffic distribution efficiency.
That shifts the core question in the earnings discussion. Instead of asking why Google is spending so heavily, the more important issue is how much revenue and profit each additional gigawatt of compute can produce.
Why Intel is the key marginal variable
Industry media had previously reported that Google plans to have Intel manufacture more than 3 million TPUs in 2028, giving Google a second source of supply beyond Taiwan Semiconductor Manufacturing Co. The report has not been confirmed by either Google or Intel, and Reuters was unable to independently verify it.
Even so, WhiteLine Daily argued that a deal of that kind would mean far more than a standard chip order if it is finalized. It would show that Intel can handle large-scale production demand from a top-tier AI customer. It would also serve as outside validation for Intel’s process technology, yields and delivery capabilities in foundry manufacturing.
The note added that Google’s interest in diversifying its supply chain would not be surprising. Demand tied to AI has kept advanced-process capacity tight, and relying only on TSMC may not be enough to support the pace of future expansion. The real uncertainty lies elsewhere: whether Intel can execute on schedule.
What the market is watching next
After Google raised its CapEx outlook, the first issue for markets is whether heavier spending can reverse semiconductor momentum, or whether it will deepen concerns around free cash flow, credit and index volatility.
From an industry perspective, though, WhiteLine Daily said Google is not pulling back on AI investment. It is locking in compute for the coming years ahead of time. As long as its credit profile and financing capacity remain far from any red line, capital spending is likely to keep climbing.
That is why Intel may be the most elastic marginal beneficiary in this setup. But the conclusion still depends on one condition. A Google order would only prove demand exists. Intel would need to deliver on time to prove it is truly back.