從幻想到現實：技術泡沫的相似之處

每當我看著投資者將數十億美元投入到另一場技術革命中時，我都會想起馬克·吐溫的智慧，“歷史不會重演”

Whenever I watch investors pour billions into yet another technological revolution, I am reminded of Mark Twain's wisdom, “history doesn't repeat itself, but it often rhymes.” His words best describe the parallels between the dot-com era and the current fever over artificial intelligence.

每當我看著投資者將數十億美元投入到另一場技術革命中時，我都會想起馬克·吐溫的智慧：“歷史不會重演，但它經常押韻。”他的話最好地描述了點-Com時代與當前對人工智能的發燒之間的相似之處。

From illusion to reality: parallels of tech bubbles

從幻想到現實：技術泡沫的相似之處

Do you remember the dot-com boom of the late ‘90s? That was an era of boundless optimism, when every company with the prefix .com in its name seemed like a gold mine.

您還記得90年代後期的互聯網繁榮嗎？那是一個無限樂觀的時代，當時每個以其名稱的前綴.com的公司似乎都是金礦。

Behind the scenes, however, a less visible but no less dramatic story was unfolding: the story of fiber optic technology, which promised to build the “information superhighway” of the future, where data would travel at the speed of light.

然而，在幕後，一個不太明顯但同樣具有戲劇性的故事正在展開：光纖技術的故事，它有望建立未來的“信息超高速”，數據將以光速傳播。

In the shadow of big names like Boo.com, another bubble was quietly inflating: fiber-optic firms that took off like rockets on promises of endless growth. Their stories are similar to today's narratives about the artificial intelligence revolution.

在Boo.com之類的大人物的陰影中，另一個泡沫正在悄悄地膨脹：光纖公司像火箭一樣起飛了無盡增長的承諾。他們的故事類似於當今關於人工智能革命的敘述。

Ciena and other companies sold the future based on Forbes' prediction that "communications bandwidth will triple every year for 25 years." Telecoms invested over $500 billion in fiber optic infrastructure that was 85-95% unused after the collapse of the bubble.

CIENA和其他公司根據《福布斯》的預測，即“通信帶寬每年將在25年內將三倍”出售未來。電信在泡沫崩潰後未使用85-95％的光纖基礎設施投資了超過5000億美元。

Today, we see an almost identical picture in artificial intelligence.

今天，我們在人工智能中看到了幾乎相同的圖片。

The investment fever of our days

我們時代的投資熱

Four tech giants - Amazon, Google, Meta and Microsoft - announced $320 billion in capital spending this year alone. That's 40% more than last year, and roughly equivalent to two-thirds of all fiber investments in the entire '90s, but spent in just one year.

四家科技巨頭 - 亞馬遜，Google，Meta和Microsoft-宣布僅今年的資本支出為3200億美元。這比去年增加了40％，大約相當於整個90年代所有光纖投資的三分之二，但僅在短短一年內就花費了。

Mark Zuckerberg isn't shy about scale - he announced the construction of a Manhattan-sized data center for artificial intelligence. The $500 billion Stargate project and the UAE's initiative to invest $1.4 trillion in U.S. artificial intelligence infrastructure suggest that we are in the midst of an unprecedented investment boom.

馬克·扎克伯格（Mark Zuckerberg）對規模並不害羞 - 他宣佈建設曼哈頓大小的人工智能數據中心。這項耗資5000億美元的星際之門項目和阿聯酋為美國人工智能基礎設施投資1.4萬億美元的倡議表明，我們正處於前所未有的投資繁榮之中。

Nvidia as a symbol of an era

nvidia是一個時代的象徵

The main beneficiary of this boom has been Nvidia. Since the introduction of ChatGPT, its market capitalization has grown from $400 billion to $2.8 trillion - a sevenfold increase! And it's not just speculative growth - the company's sales have soared from $27 billion in 2022 to $130 billion in 2024.

這種繁榮的主要受益者是Nvidia。自從引入CHATGPT以來，其市值已從4000億美元增長到2.8萬億美元 - 增長了7倍！這不僅是投機性增長 - 該公司的銷售額從2022年的270億美元飆升至2024年的1300億美元。

Nvidia CEO Jensen Huang believes that we've entered an era of "agent-based artificial intelligence" that requires 100 times more computing resources than anticipated a year ago. This he attributes to the need for massive investments in chips and data centers.

NVIDIA首席執行官詹森·黃（Jensen Huang）認為，我們進入了一個“基於代理的人工智能”時代，其計算資源的計算資源比一年前的預期要高100倍。他歸因於對芯片和數據中心進行大量投資的需求。

But even CEOs of tech giants recognize an element of FOMO (fear of lost profits) in their investment decisions. Google CEO Sundar Pichai recently stated, "At this stage of exponential growth, the risk of underinvestment is dramatically greater than the risk of overinvestment."

但是，即使是科技巨頭的首席執行官，也認識到投資決策中FOMO（擔心利潤損失）的要素。 Google首席執行官Sundar Pichai最近表示：“在指數增長的這個階段，投資不足的風險大大大於投資過度的風險。”

An alarm from the Middle Kingdom

中間王國的警報

On January 20, a little-known Chinese company DeepSeek released an artificial intelligence model R1, which is able to reason at a level comparable to OpenAI models. But the main surprise was that its training cost many times cheaper than its competitors, while using old-generation chips.

1月20日，一家鮮為人知的中國公司DeepSeek發布了人工智能Model R1，能夠以與OpenAI模型相當的水平進行推理。但主要的驚喜是，使用舊代籌碼時，其培訓的成本比競爭對手便宜多次。

That news started a chain reaction on Wall Street dubbed "Artificial Intelligence Black Monday," the largest single-day selloff of Nvidia stock in U.S. history at $600 billion.

該消息開始在華爾街（Wall Street）稱為“人工智能黑色星期一”的連鎖反應，這是美國歷史上最大的NVIDIA股票的最大單日拋售，為6000億美元。

DeepSeek has demonstrated that the future of artificial intelligence may not be in brute computing power, but in algorithmic efficiency - models learn to think and know where to serve food, rather than memorizing everything. This can radically reduce the cost of training and running models.

DeepSeek表明，人工智能的未來可能不是蠻橫的計算能力，而是在算法效率中 - 模型學會思考並知道在哪裡提供食物，而不是記住一切。這可以從根本上降低培訓和運行模型的成本。

An equation with unknown profits

利潤未知的方程式

Another troubling signal is that the artificial intelligence economy is not yet working. By some estimates, America's largest software companies have earned only $10-20 billion from artificial intelligence products this year. That's a drop in the ocean compared to nearly half a trillion dollars in investment.

另一個令人不安的信號是人工智能經濟尚未起作用。據一些估計，美國最大的軟件公司今年僅從人工智能產品中賺了100億美元。相比之下，這是海洋中的下降。

Even OpenAI, in the words of Sam Altman, isn't breaking even, despite a subscription fee of $200 a month for a premium plan.

即使用山姆·奧特曼（Sam Altman）的話說，即使openai也沒有打破，儘管每月訂閱200美元的高級計劃費用。

A story waiting to rhyme

一個等待押韻的故事

According to a sobering report by the Organisation for Economic Co-operation and Development, artificial intelligence will add just 1% to GDP growth over the next decade. While the technology could theoretically replace up to 20% of jobs, that figure shrinks to a modest 5% when real-world implementation costs are taken into account.

根據經濟合作與發展組織的一份清醒報告，人工智能將在未來十年中增加1％的GDP增長。儘管從理論上講，這項技術最多可以代替20％的就業機會，但考慮到現實世界實施成本時，該數字縮小了5％。

We are witnessing a classic situation where the ambitions of technology giants and investors are far ahead of the actual implementation of the technology. As the head of Nvidia noted, the industry was wrong in its forecasts: "The demand for computing power and scaling of artificial intelligence is not only not decreasing, but is actually growing at an incredible acceleration.

我們目睹了一個經典的情況，即技術巨頭和投資者的野心遠遠領先於技術的實際實施。正如Nvidia的負責人所指出的那樣，該行業的預測是錯誤的：“對計算能力和人工智能的規模的需求不僅不僅在減少，而且實際上以令人難以置信的加速增長。

And there's a serious threat to the

而且對