저자: 미상 · 외국인
요약 보기
이 영상은 GMO 공동창업자 Jeremy Grantham과 역사학자 Edward Chancellor가 2026년 1월 발표한 AI 거품 논문을 해설한다. 결론적으로 진행자는 미국 주식시장과 AI가 거품 단계에 있으며, 장기적으로 대다수 투자자가 큰 손실을 볼 것이라는 논문의 입장을 소개한다. 그랜섬은 2000년 1월 S&P500이 10년간 연 약 2% 실질 손실을 낼 것이라 예측했고, 실제로는 연 약 3.5% 손실이 났다고 설명한다. 그는 거품을 장기 실질 가격 추세 대비 2표준편차 이상 벗어난 상태로 정의하며, 300여 개 사례 모두 결국 추세로 복귀했다고 주장한다. 논문의 거품 10가지 신호에 대한 진행자의 채점 결과: 새 기술에 대한 과장된 주장(충족), 쉬운 돈(금리는 높지만 여전히 낮다는 해석으로 조건부 충족), 일반적 낙관(제로데이 옵션이 S&P500 옵션의 60% 이상), 관련 출판물 급증(충족), 신규 기업 양산 기계(충족), 전통 밸류에이션 중단(CAPE 약 40~40.5로 1999년 고점 44에 근접), 미성숙 기술(MIT 연구상 기업 AI 파일럿의 5%만 수익 개선), 과잉 자본 투입(아마존·알파벳·메타·마이크로소프트의 최근 12개월 설비투자 약 5,110억 달러, 3년 전 대비 3.5배), 회계 부정(서버 내용연수 연장·순환 거래로 부분 충족), 조정(아직 미발생). 진행자는 10개 중 약 8.5개가 충족된다고 본다. 논문이 제시한 정점 신호 3가지(투기주 붕괴, 우량주의 시장 초과 수익, 상승세 둔화)는 2026년 9월 말 기준 아직 나타나지 않았다고 진행자는 확인한다. 최고 투기 영역 약 9%, 우량주 약 10%, S&P500 최근 6개월 22% 상승이라는 수치를 제시한다. 진행자는 그랜섬이 과거 경고 이후에도 시장이 계속 올랐으므로 너무 이르다는 위험이 있음을 인정한다. 투자 시사점으로는 고밸류에이션 이후 저수익 경향, 기술 승자와 투자 승자의 불일치(철도·아마존 사례), 지수 집중도(JP Morgan 추정 AI 관련 30개 기업이 S&P500의 44%), 'This time is different'의 위험 등을 제시한다. 개별 종목·가격 목표에 대한 직접 발언은 없으며, 영상 말미에 Michael Burry도 메모리 반도체에 대해 유사하게 베팅 중이라는 언급이 있으나 방향 예측으로 보기는 어렵다.
다듬은 글
In January 2000, Jeremy Grantham's firm predicted that American stocks would lose money after inflation for a whole decade, and they did. This January, at 87, he published his verdict on artificial intelligence, an extreme bubble. The paper lists 10 signs that every technology mania in history has shown, from the railways of the 1840s to the internet.
I'll show you how many of those 10 AI already shows, and the one signal Grantham himself says is still missing. First, who he is and why his warnings get read. Grantham co-founded the Boston investment firm GMO in 1977. Six years earlier, he was one of the first people to recommend index funds, back when almost nobody had heard of them.
But, he's famous for bubbles. In January 2000, at the very top of the dot-com mania, his firm forecast that the S&P 500 would lose about 2% a year after inflation for 10 years. It seemed outlandish at the time, and it turned out too optimistic, because the actual loss was about 3 and 1/2% a year. The paper's comment on that is my favorite line in it, "We hadn't been pessimistic enough." It may be the only time in history an investor complained about not being gloomy enough.
His new book is called The Making of a PermaBear, and PermaBear is what his critics have long called him. It's the only book title in finance that doubles as a Wall Street insult. The paper has two parts, and the first one starts with a definition. At GMO, we have always defined a bubble as a two standard deviation divergence of the price of any asset class above its long-term real price trend.
In plain English, a price that has run further above its normal level than it does in all but the most extreme times. Grantham says his team studied more than 300 such bubbles across every kind of asset. In big developed stock markets, every single one of them eventually fell all the way back to its old trend.
1929, 1972, 2000, and American housing in 2006. Then he admits something unusual for a man with his record. After the peak at the end of 2021, stocks fell hard in 2022 just as his rule predicted. And then chat GPT arrived and in his phrase, it hit a deflating market like a multi-stage rocket. The rule from history is that great technological innovations lead to great bubbles.
The second part of the paper is by the historian Edward Chancellor who co-wrote the new book with him. To jump to the conclusion, we believe that we are in a US stock market and AI bubble and that at some stage in the future, most investors will be nursing very large losses. To get there, Chancellor walks through 10 signs of a technology mania
and I'll keep score as we go. Sign one is a new technology about which extravagant claims can be made and seem justified. When railways arrived, one British paper promised that the whole world would become one great family speaking one language. In 1904, Nikola Tesla said wireless would turn the entire earth into a huge brain.
And today, the boss of OpenAI talks about fixing the climate, colonizing space, and the discovery of all of physics. So that's a tick. Sign two is easy money. The railway mania of the 1840s took off when the Bank of England cut its rate to 2%, a record low at the time. Today, rates are higher than they were in the pandemic, but the paper argues they're still low compared with growth and inflation, and governments are running wartime size deficits.
Call that a tick with an asterisk. Sign three is general optimism, and here the paper points to a strange split. Ordinary consumers feel gloomy, but investor surveys are near record highs, margin debt is at a record, and a record share of household wealth sits in stocks. Zero-day options, used for same-day betting, now make up more than 60% of all options on the S&P 500, so that's a tick.
Sign four is a wave of new publications cheering the technology. In 1845, new railway newspapers appeared almost every week. Today, it's podcasts, newsletters, and a hundred new AI research papers every single day, so that's a tick. Sign five is a machine for creating new companies, because, as the old Wall Street saying goes, when the ducks quack, feed them.
In the South Sea Bubble of 1720, one company raised money for carrying on an undertaking of great advantage, but no one to know what it is. And the paper says a former OpenAI executive's new company was valued at $12 billion before telling investors what it was working on, so that's a tick. Sign six is when normal ways of valuing stocks get suspended, and this one I can show you with public data.
This is the cyclically adjusted price-to-earnings ratio, the CAPE, a measure made famous by the Nobel economist Robert Shiller, going back to 1881. Its long-run average is about 18. It peaked at 33 in 1929 and at 44 in December 1999, the all-time record. When the paper was written, it was about 40, and in September, it was 40 and 1/2.
So, that's a tick. Sign seven is an immature technology. The paper cites a widely quoted study from MIT that found only 5% of corporate AI pilot projects showed any measurable gain in revenue or profit. And Grantham adds his own story. When his team used AI to search scientific papers, it invented reasonable-sounding studies, complete with realistic citations.
Which, to be fair, is also how a lot of financial newsletters work. So, that's a tick. Sign eight is a huge overcommitment of capital, and this is the heart of the paper. At the height of the railway mania, Britain was pouring about 7% of its national income into railways, and at one point, there were three separate lines between London and Peterborough.
In 2000, telecom companies built fiber for internet traffic they said was doubling every 100 days, when it was really doubling every year. The paper says four giants, Amazon, Alphabet, Meta, and Microsoft, spent nearly $300 billion on buildings and equipment in 2025. So, I added up their own filings for the latest 12 months, and the number didn't slow down after the paper, it went up.
$511 billion in the 12 months to June, three and a half times what the same four companies spent three years earlier. So, that's a tick. In bold. Sign nine is what the economist John Kenneth Galbraith called the bezel, the fraud and creative accounting that grows quietly in good times. The paper is careful here, and it doesn't accuse anyone of breaking the law.
But it notes that the giants have stretched the assumed life of their servers and chips from 3 years to 5 and 1/2 or 6, which makes today's profits look bigger. And it points to circular deals where a chipmaker invests in a customer who then spends the money on that chipmaker's chips. Half a tick because the bezel, as the paper says, is only revealed in full after the fall.
Which brings us to sign 10, the shakeout, and this is the one that hasn't happened yet. When it does come, history is not gentle. After 2000, the Nasdaq fell 78% and Amazon stock fell more than 90%. And look at Cisco, briefly the most valuable company in the world in March 2000. It fell almost 90% and it only climbed back to that peak last December, a quarter of a century later.
So by my count, that's eight and a half out of 10 with the shakeout still to come. Now for the part the paper can't show you because it was written in January. In the paper, Grantham names the three things that usually mark the very top. A collapse of the most speculative stocks, a period when quality companies clearly beat the market, and a slowdown in the market's rise.
In January, he wrote that none of the three were evident yet. So I checked all three at the end of September. The most speculative corner of the market, measured by the best-known innovation fund, is up about 9% since mid-January. Quality stocks are up about 10%, a little behind the market, not ahead of it.
And the S&P 500 has risen 22% in the last 6 months, which is not what slowing down looks like. So, the checklist of a bubble is nearly full, and the exit signal still hasn't flashed. Here's my one doubt, and it's about the clock again. Grantham has been warning about American stocks for years, and the market has kept climbing through many of those warnings.
He even writes that this is the first great bubble in his study that didn't go all the way back down after its first fall. Being early is his professional hazard, and the title of his book says he knows it. So, what do you actually do with a paper like this? First, remember the most important sentence in it.
High valuations are, on average, followed by low returns and low valuations by high ones. It's not a timing tool. It's a warning about what the next 10 years may pay you. Second, remember that both things in the title can be true at once. Railways changed the world, and railway shares still fell by more than 65% after the mania.
Amazon changed the world, and its stock still lost more than 90% on the way. The technology can win while the early investors lose. Third, check how much of your money depends on one story. The paper sites a JP Morgan estimate that 30 AI-linked companies make up 44% of the S&P 500. So, an index fund is not as diversified as it looks.
And finally, the line Chancellor quotes from Sir John Templeton, engraved, he says, on the heart of every value investor. This time it's different are the four most expensive words in the English language. In an earlier video, I covered Michael Burry who is making a similar bet right now against memory chips.
Grantham doesn't tell you the date and neither does Burry. What they give you is a map of how these stories usually end and the humility to admit they might be early. Now I'd like to hear from you. Is AI a bubble, a golden era, or both? Write one word in the comments, bubble, golden, or both. And so I know how to pitch the next video, was this too simple, just right, or too deep?
Just write one word, simple, right, or deep. Thanks for watching and if you made it to the end, subscribe because next time we're opening another letter from a legend.
Comments 0