As of 6/30/26, the year-to-date returns are 10.2% return for the S&P and the 13.1% return for the NASDAQ. The year-to-date gains in both indices are heavily concentrated in just a few companies. Just ten companies (all semiconductor related) accounted for 99% of the YTD gains in the NASDAQ 100 and 78% of the gains for the S&P 500.
Such a semiconductor-led, tech-driven market has not been seen since the late 90s. Mark Twain famously said, “History doesn’t repeat itself, but it often rhymes.” Most investors alive today don’t have memories of investing through the dotcom bubble. Read the following comparison of 1999 and 2026 and see if it doesn’t give you a sense of déjà vu.
1999
2026
The bull market was driven by a revolutionary technology that was blindingly obvious to all – the internet.
The bull market is driven by a revolutionary technology that is blindingly obvious to all – artificial intelligence.
There was a massive, unprecedented investment in infrastructure in laying fiber optic cables and building telecommunications routing and networking hardware.
There is a massive, unprecedent investment in infrastructure in building data centers and installing AI compute chips.
Cisco, the “pick and shovel” provider of routers and switches used in the infrastructure build, was massively profitable and became the most valuable company in the world.
Nvidia, the “pick and shovel” provider of AI compute chips used in data centers, is massively profitable and became the most valuable company in the world.
Cisco had a moat in IOS (Internetwork Operating System) that ran its routers and switches. Millions of network engineers were trained on Cisco certifications. Shifting to a competitor meant completely retraining an entire IT department and rewriting thousands of lines of network configuration architecture.
Nvidia has a moat in CUDA (Compute Unified Device Architecture). Many AI researchers and developers were trained on CUDA. Shifting to competitor hardware often requires developers to completely rewrite complex code, re-validate models, and fix bugs without the massive online community support that CUDA enjoys.
Cisco used circular funding and lent billions to cash-strapped startup telecom providers who used that cash to buy Cisco routers.
Nvidia uses circular funding and invested billions in cash-strapped neoclouds like CoreWeave and Lambda Labs who used that cash to buy Nvidia GPUs.
Semiconductor chip companies such as Micron and Sandisk soared hundreds and even thousands of percent.
Semiconductor chip companies such as Micron and Sandisk soared hundreds and even thousands of percent.
Importantly, the end customers of Global Crossing, WorldCom, the last-mile telecom companies, and the ecommerce darlings of Webvan and Pets.com were all unprofitable.
Importantly, the end customers of Coreweave, Lambda Labs, and the AI labs of OpenAI, Anthropic, and Grok are all unprofitable.
With many internet companies losing money, investors measured progress by focusing on new metrics such as users and page views.
With many AI companies losing money, investors measure progress by focusing on new metrics such as users and tokens consumed.
Investors adopted a gambling mentality, and day trading became popular. Speculators heavily employed the use of leverage through margin loans.
Investors adopted a gambling mentality, giving rise to increased retail trading, meme stocks, and betting markets. Speculators heavily employed the use of leverage through options contracts and leveraged ETFs.
The speculative fervor allowed companies to inflate their stocks just by adding .com to their names.
Stock of failed shoe company Allbirds soared nearly 600% in 1 day after pivoting its business to AI compute infrastructure, later renaming itself Smartbird.
The Spurs and the Knicks faced off in the NBA finals.
The Spurs and the Knicks faced off in the NBA finals. (Okay, this has nothing to do with markets, but kind of amusing).
The Schiller PE ratio first reached 42 in 1999 before peaking at 44 in 2000.
The Schiller PE ratio recently touched 42.
There was a flood of IPOs and the increased supply of stock being sold into the market may have contributed to the end of the bull market.
We are finally seeing a surge in IPOs with SpaceX, OpenAI, and Anthropic all trying to go public in 2026. There could be a surge in selling pressure as the insider lockups end 6-12 months post IPO.
The NASDAQ peaked in March of 2000 without a clear catalyst. Time, overvaluation, and hype simply ran its course.
To be determined…
In the aftermath, many of the tech darlings lost 70-90% of their value.
To be determined…
What happened in 1999 wasn’t just that investors chased internet stocks. Investors also abandoned boring, old-economy stocks. I find it forever amusing that the day NASDAQ hit its peak on 3/10/2000 was the exact same day that Berkshire Hathaway hit its multi-year low. From the beginning of 1999 through March 2000, NASDAQ rose 130% while Berkshire fell 40%.
The fact that many internet companies lost money did not matter. The fact that Berkshire was very profitable and had a phenomenal track record did not matter. All investors cared about was owning a piece of the ever-rising internet and semiconductor gravy train. Of course, the following two years could not have been more different as Berkshire rose 80% while NASDAQ declined 60%. Such is the peril of chasing the hot trend and investing through the rear-view mirror.
This phenomenon wasn’t unique to Berkshire Hathaway. The following shows the performance of tech and semiconductors stocks before and after the bubble burst on the left-hand side and the performance of the old-economy stocks on the right-hand side. Note how the profitable, old economy stocks underperformed as the bubble inflated and outperformed as the bubble deflated.
A similar dynamic is occurring this year in both the US and Asia. In the US, profitable, non-AI stocks such as Berkshire, Markel, and Disney are down as capital flows into AI and semiconductor companies. In Asia, giants such as Tencent, Alibaba, and JD.com are all down as investor sell old tech stocks to chase the new AI stocks in Hong Kong (such as Knowledge Atlas) and Korean semiconductor stocks (Samsung, SK Hynix). The valuation for Knowledge Atlas, a Chinese AI company focused on developing large language models, is crazy at $120 billion dollars despite having just $100 million in revenues and $500 million in losses in 2025. There is no way this company should be worth half of Alibaba or a quarter of Tencent. I continue to believe that our portfolio is exactly what you want to own when the AI bubble eventually bursts.
A fair question you might ask is what if you are wrong? What if AI is different? What if this time is different? Even though “this time is different” are the four most dangerous words in investing, it is important to keep an open mind. I am constantly looking for evidence to reevaluate whether AI is generating a good return on capital for the end users. So far, the evidence is mostly negative.
ByteDance
ByteDance is the owner of TikTok in the US and Douyin in China. It is also the leading consumer AI company in China through Doubao, the #1 consumer AI app in China. Despite the strong consumer adoption of Doubao, Forbes reported ByteDance has cut roughly 30% of its AI applications projects and explicitly abandoned the “spray-and-pray” product strategy that once defined China’s mobile internet playbook. ByteDance’s AI inference costs in 2025 reportedly exceeded RMB 8 billion – approximately 2.3 times its incremental revenue from AI products.
In the mobile era, scale drove profitability; marginal costs approached zero as user growth accelerated. In AI, every additional query incurs real compute and storage costs — unlike the mobile internet era, where costs were largely offloaded to user devices, AI centralizes that burden back onto the provider. Growth deepens losses.
SpaceX
This quarter we had the largest IPO in history in SpaceX. Importantly, we can now start to track the reported financials for xAI (Grok). SpaceX reports its AI segment (the old Twitter plus Grok) separately from its space and connectivity businesses. As you can see below, the AI segment is reporting modest revenue growth with large losses. Operating losses have ballooned from 60% of sales in 2024 to 300% of sales in the first quarter of 2026. Actual cash burn is even greater as SpaceX spent almost $8 billion on AI capital expenditures in the latest quarter (which only gradually show up as depreciation expense over many years).
SpaceX is valued at an incredible $2 trillion dollar market cap despite having just $20 billion of revenues. The company’s own IPO prospectus attributes 90%+ of the total addressable market to the AI business. Yet Grok is at best maybe fourth place in the AI race behind Anthropic, Gemini, and OpenAI.
Today, SpaceX’s price is supported by having less than 5% of its shares outstanding available for trading. Over the next six months, another 35% of shares held by insiders and early investors will be gradually released from lock-up. This steady increase in share count could create additional selling pressure on the stock.
OpenAI
We’ll soon also be able to see OpenAI’s financials as it is pursuing an IPO. Meanwhile, we can track ChatGPT’s progress by looking at Google search traffic. The deceleration in the growth of ChatGPT has continued with search traffic nearly flat year over year for the second quarter. There is a high likelihood ChatGPT will see declines in the second half of the year as it loses share to Gemini and Claude.
Wall Street Journal reported that OpenAI missed an internal goal of reaching one billion weekly active users for ChatGPT by the end of last year. The company still hasn’t announced that milestone, unnerving some investors. It also missed its yearly revenue target for ChatGPT as well after Google’s Gemini saw massive growth late last year and ate into OpenAI’s market share, the people said. Missing growth targets is problematic for OpenAI because it is bleeding money and depends on investor confidence to raise money to fund its cash burn. Chief Financial Officer Sarah Friar has told other company leaders that she is worried the company might not be able to pay for future computing contracts if revenue doesn’t grow fast enough.
Tokenmaxxing Backfires
The one bright spot in AI this year has been the strong adoption of Anthropic’s Claude by businesses. Some of this may have been driven by tokenmaxxing, the corporate practice of grading employees on token usage (often without adequate focus on output). This predictably backfired as AI token usage costs spiraled out of control at many businesses without offsetting increases in productivity. Employees apparently deliberately inflated their AI usage on low value tasks to climb the token leaderboards
Microsoft cancelled their Claude Code pilot after token-based billing consumed the team’s entire annual budget within months.
Uber CTO Praveen Neppalli Naga disclosed in April that the company had already exhausted the Claude Code budget allocated for all of 2026. Uber’s COO publicly noted that higher token consumption has not translated into a proportional increase in useful, consumer-facing features. Moving forward, the company is enforcing strict calculations to tie AI spend directly to shipped features.
Amazon has shut down an internal leaderboard that tracked employees’ use of AI tools after workers tried to boost their scores with unnecessary activity that increased the company’s computing costs. Dave Treadwell, an Amazon senior vice-president, told staff, “Please don’t use AI just for the sake of using AI.”
Meta Chief Technology Officer Andrew Bosworth said in an April memo to employees. “It has been great to let people experiment but now we have too many overlapping tools. Nobody should be using AI tools just for the sake of using them. All motion is not progress, and token usage alone is not a measure of impact of any kind.”
It seems the late Charlie Munger was right, “Show me the incentive and I’ll show you the outcome.”
Disclosure
I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it.
The information in the article is provided for informational purposes only. It should not be construed as investment advice or advice on buying, selling, or other types of transactions relating to an investment in products or services, much less an invitation, an offer or a solicitation to invest.
The information in the article is provided solely by virtue of the fact that everyone will independently make their own investment decisions: the report does not take into account investment objectives, nor specific needs or financial situation. In addition, nothing in the article represents or is intended to express investment, financial, legal, accounting or tax advice. You should consult your own investment or financial advisor before taking any actions.
While evaluating an investment manager’s historical track record is important, it is even more important to understand his or her investment philosophy. Understanding the investment
Shades of 1999
As of 6/30/26, the year-to-date returns are 10.2% return for the S&P and the 13.1% return for the NASDAQ. The year-to-date gains in both indices are heavily concentrated in just a few companies. Just ten companies (all semiconductor related) accounted for 99% of the YTD gains in the NASDAQ 100 and 78% of the gains for the S&P 500.
Such a semiconductor-led, tech-driven market has not been seen since the late 90s. Mark Twain famously said, “History doesn’t repeat itself, but it often rhymes.” Most investors alive today don’t have memories of investing through the dotcom bubble. Read the following comparison of 1999 and 2026 and see if it doesn’t give you a sense of déjà vu.
What happened in 1999 wasn’t just that investors chased internet stocks. Investors also abandoned boring, old-economy stocks. I find it forever amusing that the day NASDAQ hit its peak on 3/10/2000 was the exact same day that Berkshire Hathaway hit its multi-year low. From the beginning of 1999 through March 2000, NASDAQ rose 130% while Berkshire fell 40%.
The fact that many internet companies lost money did not matter. The fact that Berkshire was very profitable and had a phenomenal track record did not matter. All investors cared about was owning a piece of the ever-rising internet and semiconductor gravy train. Of course, the following two years could not have been more different as Berkshire rose 80% while NASDAQ declined 60%. Such is the peril of chasing the hot trend and investing through the rear-view mirror.
This phenomenon wasn’t unique to Berkshire Hathaway. The following shows the performance of tech and semiconductors stocks before and after the bubble burst on the left-hand side and the performance of the old-economy stocks on the right-hand side. Note how the profitable, old economy stocks underperformed as the bubble inflated and outperformed as the bubble deflated.
A similar dynamic is occurring this year in both the US and Asia. In the US, profitable, non-AI stocks such as Berkshire, Markel, and Disney are down as capital flows into AI and semiconductor companies. In Asia, giants such as Tencent, Alibaba, and JD.com are all down as investor sell old tech stocks to chase the new AI stocks in Hong Kong (such as Knowledge Atlas) and Korean semiconductor stocks (Samsung, SK Hynix). The valuation for Knowledge Atlas, a Chinese AI company focused on developing large language models, is crazy at $120 billion dollars despite having just $100 million in revenues and $500 million in losses in 2025. There is no way this company should be worth half of Alibaba or a quarter of Tencent. I continue to believe that our portfolio is exactly what you want to own when the AI bubble eventually bursts.
A fair question you might ask is what if you are wrong? What if AI is different? What if this time is different? Even though “this time is different” are the four most dangerous words in investing, it is important to keep an open mind. I am constantly looking for evidence to reevaluate whether AI is generating a good return on capital for the end users. So far, the evidence is mostly negative.
ByteDance
ByteDance is the owner of TikTok in the US and Douyin in China. It is also the leading consumer AI company in China through Doubao, the #1 consumer AI app in China. Despite the strong consumer adoption of Doubao, Forbes reported ByteDance has cut roughly 30% of its AI applications projects and explicitly abandoned the “spray-and-pray” product strategy that once defined China’s mobile internet playbook. ByteDance’s AI inference costs in 2025 reportedly exceeded RMB 8 billion – approximately 2.3 times its incremental revenue from AI products.
In the mobile era, scale drove profitability; marginal costs approached zero as user growth accelerated. In AI, every additional query incurs real compute and storage costs — unlike the mobile internet era, where costs were largely offloaded to user devices, AI centralizes that burden back onto the provider. Growth deepens losses.
SpaceX
This quarter we had the largest IPO in history in SpaceX. Importantly, we can now start to track the reported financials for xAI (Grok). SpaceX reports its AI segment (the old Twitter plus Grok) separately from its space and connectivity businesses. As you can see below, the AI segment is reporting modest revenue growth with large losses. Operating losses have ballooned from 60% of sales in 2024 to 300% of sales in the first quarter of 2026. Actual cash burn is even greater as SpaceX spent almost $8 billion on AI capital expenditures in the latest quarter (which only gradually show up as depreciation expense over many years).
SpaceX is valued at an incredible $2 trillion dollar market cap despite having just $20 billion of revenues. The company’s own IPO prospectus attributes 90%+ of the total addressable market to the AI business. Yet Grok is at best maybe fourth place in the AI race behind Anthropic, Gemini, and OpenAI.
Today, SpaceX’s price is supported by having less than 5% of its shares outstanding available for trading. Over the next six months, another 35% of shares held by insiders and early investors will be gradually released from lock-up. This steady increase in share count could create additional selling pressure on the stock.
OpenAI
We’ll soon also be able to see OpenAI’s financials as it is pursuing an IPO. Meanwhile, we can track ChatGPT’s progress by looking at Google search traffic. The deceleration in the growth of ChatGPT has continued with search traffic nearly flat year over year for the second quarter. There is a high likelihood ChatGPT will see declines in the second half of the year as it loses share to Gemini and Claude.
Wall Street Journal reported that OpenAI missed an internal goal of reaching one billion weekly active users for ChatGPT by the end of last year. The company still hasn’t announced that milestone, unnerving some investors. It also missed its yearly revenue target for ChatGPT as well after Google’s Gemini saw massive growth late last year and ate into OpenAI’s market share, the people said. Missing growth targets is problematic for OpenAI because it is bleeding money and depends on investor confidence to raise money to fund its cash burn. Chief Financial Officer Sarah Friar has told other company leaders that she is worried the company might not be able to pay for future computing contracts if revenue doesn’t grow fast enough.
Tokenmaxxing Backfires
The one bright spot in AI this year has been the strong adoption of Anthropic’s Claude by businesses. Some of this may have been driven by tokenmaxxing, the corporate practice of grading employees on token usage (often without adequate focus on output). This predictably backfired as AI token usage costs spiraled out of control at many businesses without offsetting increases in productivity. Employees apparently deliberately inflated their AI usage on low value tasks to climb the token leaderboards
Microsoft cancelled their Claude Code pilot after token-based billing consumed the team’s entire annual budget within months.
Uber CTO Praveen Neppalli Naga disclosed in April that the company had already exhausted the Claude Code budget allocated for all of 2026. Uber’s COO publicly noted that higher token consumption has not translated into a proportional increase in useful, consumer-facing features. Moving forward, the company is enforcing strict calculations to tie AI spend directly to shipped features.
Amazon has shut down an internal leaderboard that tracked employees’ use of AI tools after workers tried to boost their scores with unnecessary activity that increased the company’s computing costs. Dave Treadwell, an Amazon senior vice-president, told staff, “Please don’t use AI just for the sake of using AI.”
Meta Chief Technology Officer Andrew Bosworth said in an April memo to employees. “It has been great to let people experiment but now we have too many overlapping tools. Nobody should be using AI tools just for the sake of using them. All motion is not progress, and token usage alone is not a measure of impact of any kind.”
It seems the late Charlie Munger was right, “Show me the incentive and I’ll show you the outcome.”
Disclosure
I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it.
The information in the article is provided for informational purposes only. It should not be construed as investment advice or advice on buying, selling, or other types of transactions relating to an investment in products or services, much less an invitation, an offer or a solicitation to invest.
The information in the article is provided solely by virtue of the fact that everyone will independently make their own investment decisions: the report does not take into account investment objectives, nor specific needs or financial situation. In addition, nothing in the article represents or is intended to express investment, financial, legal, accounting or tax advice. You should consult your own investment or financial advisor before taking any actions.
Subscribe to our newsletter
Recent Posts
Market Outlook: 2025 Review and Beyond
How does Trump’s One Big Beautiful Bill Impact You?
4 Key Parts of My Investment Philosophy
NetEase (NTES)
Find these articles helpful? Subscribe to our newsletter!
Get our latest personal finance tips and market outlook straight to your inbox!
Plus, receive our free article on “The Magic of Compounding and the Three Parts to Building Wealth.”
You Might Also Enjoy
Market Outlook: 2025 Review and Beyond
The market continued its march to new highs with the S&P 500 up 3% in the fourth quarter and up 18% for the year. Incredibly,
How does Trump’s One Big Beautiful Bill Impact You?
Trump’s One Big Beautiful Bill (OBBB) is the most important piece of tax legislation since the Tax Cuts and Jobs Act from 2017. Almost everyone
4 Key Parts of My Investment Philosophy
While evaluating an investment manager’s historical track record is important, it is even more important to understand his or her investment philosophy. Understanding the investment