“Hundreds of billions” is the polite headline. Bloomberg’s number, as carried by the Financial Post on August 27, is $692 billion of market value gone since Alphabet’s May 13 all-time high, a 15% drawdown, making Google’s parent the second-biggest point drag on the S&P 500 over that span. Cap-destruction of that size is a positioning event. It is not automatically a bargain, and it is not a Protected Wheel entry. The question the tape is asking is not whether Google has TPUs. It is whether the AI spend is producing a model lead, or just a depreciation schedule.
What Bloomberg actually reported
For much of the prior year Alphabet was the Magnificent Seven name to beat. Shares hit that May 13 high after soaring more than 150% in the previous 12 months, among the 25 best S&P 500 performers over that stretch and far ahead of the other Mag 7 names. The reversal, Bloomberg wrote, came from a brain drain at Google and fear the company is losing its AI edge — plus the delayed Gemini flagship and the capex bill.
On personnel: two senior people left for Anthropic and OpenAI a few months earlier. In August, Jeff Dean — “key to Google’s AI strategy,” in Bloomberg’s phrasing — left to launch a startup and took high-profile coworkers with him. Demis Hassabis stepped down as CEO of Google DeepMind and became chairman. Those moves sent shares down 4% on August 5, erasing $186 billion in a session.
On the model: Gemini 3.5 Pro is behind schedule as the company works on coding, an area Bloomberg said Alphabet is already seen as lagging Anthropic and OpenAI. Two weeks before the August 27 piece, Google released Gemini 3.7 Flash and gave no timetable for 3.5 Pro. A Google spokesperson’s pushback, quoted in the same article: “Our AI momentum and shipping velocity are at an all-time high. We’re rolling out model updates within weeks of each other, with Gemini 3.7 Flash becoming our fastest-growing model to date and Gemma surpassing one billion downloads.”
On the spend: Alphabet raised $25 billion in an early August bond offering. Investors have been pressing hyperscalers for proof of return on data-center build-out. The Nasdaq 100, Bloomberg noted, had not made a new high since May. Alphabet sold off after late-July earnings on heavy capex and negative free cash flow. Janus Henderson analyst Divyaunsh Divatia, quoted there: capex “is going to go up significantly next year,” free cash flow stays negative, and the company has already raised equity and a lot of debt to fund “this race to AGI.” That makes a lot of investors nervous, he said.
Separately, Yahoo Finance and company reports put second-quarter capex at $44.9 billion, full-year 2026 guidance at $195–$205 billion (up from $180–$190 billion), and free cash flow at negative $5.9 billion — Alphabet’s first negative FCF quarter since the 2004 IPO. CFO Anat Ashkenazi said FCF would stay under pressure. Those figures are from the earnings cycle, not from Bloomberg’s August 27 wrap, but they are the numbers the $692 billion drawdown is sitting on. I am not going to invent a one-day wipeout figure the August 27 Bloomberg piece did not use; some later recaps cited ~$293 billion on the post-earnings session. Treat session-level cap-loss figures as recap math unless you pull the tape yourself.
Despite the slump, Bloomberg/Financial Post said the stock was still up 65% over the trailing 12 months and had outperformed Mag 7 peers over that window. Business Insider, the same week, had shares around $336 on Thursday, still up 7% year to date. Do not confuse a 15% drawdown from the high with a destroyed franchise.
What it means for a California holder
Alphabet is a Mountain View company. California 401(k)s and RSUs are already long GOOGL. Adding more because “$692 billion” sounds cheap is how you double a sector bet you already have. California taxes the gain when you sell. It does not rebate the 15% you already gave back.
CFRA’s Angelo Zino, quoted by Bloomberg: “You’ve had this kind of brain drain. It does pose some risk because it’s the area of the market everybody is looking at at this point in time, right? It’s, you know, can you monetize AI?” Visible Alpha’s Melissa Otto was blunter: she cares less about the exact person and more about what management says on the next earnings call about AI translating into fundamental growth. “That’s somewhat of an open question.”
Sands Capital’s Daniel Pilling, a holder, floated the opposite: maybe new people are what Google needs, and the TPU fleet is still among the largest compute bases in the world. That is an owner talking his book. Fine. It is not your entry rule.
Caveats
Talent wars are industry-wide. Bloomberg itself noted Meta’s packages, OpenAI’s hiring out of Apple, Apple suing OpenAI. Alphabet is not uniquely messy. It is uniquely expensive to be messy while spending on the order of $200 billion a year of capex and printing negative FCF.
A 15% drawdown after a 150% run is also just mean reversion. Microsoft, in the same Bloomberg piece, had surged more than 25% since the end of July on faster cloud growth. Rotation inside Mag 7 is not a thesis that Google is finished.
If the four entry rules on concentration, trend, liquidity, and defined risk are not met, there is no trade. This is educational commentary, not investment advice. “$692 billion gone” is a headline. The business still throws off enormous operating cash. The tape is asking whether that cash is being converted into an AI product people pay for, or into a bond prospectus.
Source: Alphabet Stock Sheds $692 Billion as AI Delays and Talent Loss Hit Shares (Carmen Reinicke, Bloomberg, Aug. 27, 2026); readable syndication: Financial Post.
The Hedge 