AI: Google Q2 leads AI Capex Ramps, Anthropic/OpenAI vs China & more. AI-RTZ #1158

1. Google Earnings & Rising AI Capex: Alphabet/Google reported strong Q2 results, and a higher AI capex budget of $205 billion. As I broke down on ARD #125, Google is ‘backing up the truck’ on AI infrastructure: it lifted the 2026 capex guide to ~$205B (from ~$190B+) even as free cash flow went negative for the first time in decades — the FT clocked a ~$6B quarterly burn — from a company still throwing off $112B+ in profits, now bolstering its balance sheet with an $80B+ equity raise, with the cloud backlog ballooning to $514B. My take: Google’s edge isn’t the front-of-line benchmark model but a slate of fast, cheap ‘good enough’ Gemini models at mass-market scale — the Internet-Explorer-vs-Netscape playbook — leaving it the best-positioned global consumer-AI company alongside Apple (not stock advice). More here and here.

   

2. More AI Capex Ramps next week: Expect other big tech companies Amazon, Meta and Microsoft to follow suit next week with their quarterly results and likely commitments to higher AI capex budgets. As I noted on ARD #125, I’d expect every one of them to be just as committed to AI capex as Alphabet — no one is ready to ‘blink,’ which is exactly the tell the market is watching for, with the NYT’s DealBook framing Google as setting the pace and the rest following. My take: with markets still flashing green lights, the risk/reward keeps favoring stepping on the gas — the moment one giant signals a pullback (and its stock rises) is the signal that would ripple through Nvidia, memory and the whole infrastructure complex. I’m not expecting it yet. More here and here.

   

3. AI Big Tech Debt > $1.65 trillion: Nikkei Asia put out a report totaling the AI debt raised on and off balance sheet by key US big tech companies, and the total comes to $1.65 trillion. This includes Google, Microsoft, Amazon, Meta and Oracle. As I dug into on ARD #125, about $1.35T of that sits on the actual balance sheets, with ~$300B+ off them — Meta’s off-balance-sheet piece alone pegged near $420B, roughly 3x its recorded debt, largely through SPVs (special-purpose vehicles). My take: it’s framed as ‘hidden and scary,’ but those SPVs are well-understood institutional vehicles — the kind that BlackRock and the big PE/VC funds run all the time — not the 2008 CDO shenanigans the comparison implies. The one real ‘08 echo I’d flag is the NASDAQ fast-tracking SpaceXAI’s (and maybe Anthropic’s/OpenAI’s) IPO into the index in weeks. For now, investors are still calculating the risk/reward as positive. More here and here.

   

4. Anthropic & OpenAI side against China open-source AIs: The two leading US frontier AI model companies, Anthropic and OpenAI, joined ranks against China’s rapid ramp of competitive open source AI models. And garnering US government attention and possible help. Despite the rest of the US Tech industry beginning to oppose the moves. As I covered on ARD #125’s companion, the two labs are effectively enlisting Washington as the ‘bad cop’ — walling US businesses off from Chinese open-source models just ahead of their own mega-AI IPOs. My take: this is the wrong move. Open source has been the substrate of every tech wave, and blocking the best of it — increasingly coming from China — mainly handicaps the US customers trying to serve the world with the best software, which is exactly why much of the rest of Silicon Valley is now pushing back. More here and here.

   

5. AMD ramps up Helios vs Nvidia: Lots of coverage this week on AMD’s launch of Helios, its first AI Data Center GPU system offering against Nvidia’s Vera Rubin and earlier Blackwell systems. Helios is AMD’s first rack-scale system — a full data-center GPU platform rather than a single chip — and it arrived with Microsoft as a marquee first buyer. As I discussed around ARD #124, it dovetails with Anthropic’s $10B+ AMD partnership (with AMD’s circular ~$5B investment back into Anthropic) as the frontier labs actively cultivate a credible second source of AI GPUs. My take: it’s healthy for the whole ecosystem — real competition on price and supply — even if everyone stays heavily Nvidia-dependent for now; it’s another sign the AI-infrastructure buildout has room for more than one arms supplier. More here.

   

Additional Items:

1. Google under fire again for ‘breaking the internet’. More here.

2. US/China get ready for second meeting at White House 9/24/26. More here and here.

(Additional Note: AI Ramblings is now a weekday Daily podcast called AI Ramblings Daily (ARD). Different content than AI-Reset to Zero (AI-RTZ), which remains a daily morning substack with now over 1150+ ‘MY TAKES’ on key AI events and issues turbulently flowing by. AI Ramblings Daily is typically a 20 minute afternoon podcast on my take on additional AI developments of the day. Both daily substack and podcasts typically discuss different AI issues and items. And are free to subscribe. Try this week’s series with ARD Episodes # 122 — Haste Makes Waste, 123 — Setting the Stage on US-China, 124 — Damn the Torpedoes, 125 — Backing Up the Truck, and #126 Threading 3 Needles here.):

Up next, the Sunday ‘The Bigger Picture’ tomorrow. Stay tuned.

(NOTE: The discussions here are for information purposes only, and not meant as investment advice at any time. Thanks for joining us here)