How the plop-o-meter works
The meter has two halves: what the news says and what AI stocks do. Each half gives a push – up when the bubble strains, down when money still flows. The news can push up to +54 or down to −25; the market up to ±25. The two pushes are added and turned into the meter’s % on a soft S-curve: no push is 40%, and every extra point of push matters a bit less the closer the meter gets to 0 or 100. So a bad week and a terrible month stay apart instead of both hitting the ceiling.
| Push | -40 | -20 | -10 | 0 | +10 | +20 | +30 | +40 | +60 |
|---|---|---|---|---|---|---|---|---|---|
| Meter | 8% | 20% | 29% | 40% | 52% | 64% | 75% | 83% | 93% |
The news half
Every six hours we take the AI stories of the last 14 days – from the paper’s own feeds plus the markets desks listed below, money news first – and Claude checks them against twelve signals. Every signal is a yes or a no, answered strictly from the stories: a “yes” needs a story behind it (linked on the meter page), and when no story says it, the answer is no. Every “letting air out” signal pushes up, every “pumping air in” signal pushes down – by its weight times how strongly it is reported:
- How many outlets. One outlet counts 50% of the weight, two 75%, three or more the full weight. A single headline shouldn’t move the meter as much as news everyone reports.
- How fresh. The newest report counts fully for 3 days, then fades to 50% by the end of the 14-day window.
| Signal | Points | Yes when | No when |
|---|---|---|---|
| AI stocks tank Do the stories report a sharp sell-off in AI stocks? | +8 | Nvidia, AMD, Broadcom, CoreWeave, Oracle or other AI names drop sharply; market value wiped out; an AI-led market slide. | Normal daily moves, a single small company's stock. |
| Debt gets nervous Is there stress in the money borrowed to build AI? | +8 | Worries about data-center debt, private credit or bond deals for AI build-outs; lenders pulling back; rising borrowing costs or credit-default bets on AI companies. | A routine bond sale that went smoothly. |
| Spending gets cut Are AI build-out plans being cut, delayed or pushed back on? | +8 | A big tech company, lab or cloud cuts, pauses or cancels data-center or GPU spending; investors punish a company for spending too much; projects stall for lack of power or money. | Spending plans are raised or confirmed. |
| It doesn't pay off Do stories say AI isn't making the money it costs? | +7 | Revenue far behind spending, big losses at AI labs, enterprise pilots failing, adoption stalling, studies finding no productivity gain, steep price cuts to win users. | Opinion pieces only speculating about it. |
| Startups wobble Are AI companies in trouble? | +7 | Down rounds, failed raises, shutdowns, fire-sale acqui-hires, layoffs at AI companies, a well-funded AI startup missing its targets. | Normal acquisitions at a good price, routine hiring changes. |
| Money goes in circles Is money going round in circles between AI companies? | +6 | Chipmakers or clouds investing in customers who then buy their chips or compute; compute-for-equity deals; reports questioning round-tripped revenue. | Ordinary customer contracts paid in cash. |
| Too many GPUs Are there signs of too much AI capacity? | +5 | GPU rental prices falling, idle or half-empty data centers, cancelled chip orders, cheaper models making the build-out look oversized. | Shortages, waiting lists, sold-out capacity. |
| Grown-ups say “bubble” Do prominent people or institutions warn of an AI bubble? | +5 | CEOs, well-known investors, central banks, the IMF or big banks say AI valuations are a bubble or overheated. | Anonymous commenters, a blogger's hunch. |
| Results beat Do AI companies report strong results? | −8 | AI revenue or earnings beat expectations, demand outstrips supply, a lab's revenue grows fast and is reported with numbers. | Promises and forecasts without results. |
| Spending goes up Is AI spending being raised? | −6 | Big tech raises its AI capex guidance, new multi-billion data-center or chip deals are signed. | Plans are only repeated, not raised. |
| Money keeps pouring in Are investors still paying up for AI? | −6 | New funding rounds at higher valuations, oversubscribed raises, big AI IPOs that go well. | Small seed rounds, flat rounds. |
| People actually pay Is there hard evidence of paid AI use at scale? | −5 | Reported numbers on paying customers, usage or measured productivity gains at large companies. | Product launches, demos and announcements without usage numbers. |
The market half
Five gauges, read from prices only. The AI basket is 16 stocks with equal weight: Nvidia, AMD, Broadcom, TSMC, ASML, Micron, Arm, Microsoft, Alphabet, Amazon, Meta, Oracle, CoreWeave, Supermicro, Vertiv, Palantir. Falling prices push the meter up, rising ones pull it down; every gauge is capped. Together the gauges range from -28 to +48; that sum is then softly limited to ±25, so prices – which swing a lot day to day – can’t outvote the news.
| Gauge | Rule | Range |
|---|---|---|
| AI stocks, past month Equal-weight change of the AI basket over 30 days. | +1 point for every 1% it fell, −1 for every 1% it rose | -12 … +12 |
| AI stocks, past week Equal-weight change of the AI basket over 7 days – catches a sudden slide. | +1.5 points for every 1% it fell, −1.5 for every 1% it rose | -8 … +10 |
| Off the highs How far the average AI stock sits below its own 3-month high. | +0.6 points per 1% below the high, beyond the first 5% | 0 … +12 |
| Chips vs. the market The chip index (SOX) against the Nasdaq over the past month – is the AI trade lagging the rest? | +0.5 points per point the chips lag, −0.5 per point they lead | -5 … +6 |
| Fear gauge The VIX, Wall Street’s volatility index. Calm is around 15, panic above 30. | +0.5 points per VIX point above 18, −0.5 per point below | -3 … +8 |
History
Every reading is kept and drawn in the chart on the meter page. Days without one (up to 28 days back) are read afterwards: only from the stories that were published by then, and with that day’s share prices – so the line shows what the news said at the time, not what we know now. When the formula changes, every stored reading is scored again with it, so the whole line is always on one scale. The bold line is the average of the past seven days; the thin one shows every single reading.
The levels
- Still inflating · 0–24
- Stretchy · 25–44
- Wobbling · 45–64
- Hissing · 65–79
- Plop imminent · 80–99
Extra sources
On top of the paper’s own feeds, the meter reads these markets desks. They don’t appear in the paper; general markets feeds only count when the story is about AI. Prices come from Yahoo Finance.
- Yahoo Finance · AI chips – Headlines on the chip and platform names in the AI basket.
- Yahoo Finance · AI infrastructure – Headlines on clouds, data-center suppliers and AI software in the basket.
- Bloomberg Markets – Markets news, incl. AI stocks and credit.
- Financial Times · Technology – Business of big tech and AI, with the money angle.
- Financial Times · Markets – Equity and credit markets.
- CNBC Finance – Wall Street and earnings.
- MarketWatch – Market-moving stories of the day.
- Fortune · Tech – AI business and funding.
- Axios – Short takes on deals, policy and markets.
- Business Insider – AI companies, layoffs and money.
- The Register · AI – Sceptical coverage of AI vendors and spending.
- The Economist · Finance – Bubble talk from the grown-ups.
It is a joke with a method: it measures how the news and the market feel, not whether or when anything will pop. Nothing here is financial advice.