This analysis is based on discrete discussions among Long Angle community members in our private forum and represents a summary of various viewpoints and experiences shared within our network.
Important Disclosures: This content is provided for educational and informational purposes only and should not be construed as legal, tax, investment, or financial advice. The opinions and strategies discussed reflect individual member perspectives and do not constitute recommendations to buy, sell, or hold any securities or investments. All investment decisions carry risk, including the potential loss of principal, and readers should conduct their own thorough research and consult with qualified financial, legal, and tax professionals before making any investment decisions. Past performance and current market discussions do not guarantee future results.
Artificial intelligence has gone from a futuristic concept to a defining force of today's markets. Over the last two years, large language models (LLMs)—the engines behind ChatGPT, Claude, Gemini, Perplexity, and others—have transformed how people code, search for information, and interact with technology.
With this explosion of demand, investors are racing to gain exposure. Perplexity, the independent AI startup, was recently valued at $20 billion. Anthropic ($183 billion valuation) secured billions in backing from Amazon and Google. OpenAI (~$500 billion) has rolled out ChatGPT-5 to hundreds of millions of users.
But here's the dilemma: investing in AI LLM companies is easier said than done. Access is limited, valuations can be frothy, and the landscape is changing so quickly that today's front-runner could be tomorrow's relic.
Inside Long Angle's community, some investors compared the moment to the search engine wars of the early 2000s, when Yahoo seemed untouchable, until Google took over.
This article explores the discussion happening around LLM investing inside Long Angle: the bull case, the bear case, and whether the real money may lie not in the models themselves but in the infrastructure powering them.
2026 High-Net-Worth Asset Allocation Report
Why Investors Are Excited About LLM Companies
Let’s start with the bullish view. Long Angle members chasing deals in OpenAI, Anthropic, or Perplexity point to four key drivers:
- Explosive user adoption: ChatGPT became the fastest consumer application to reach 100 million users. Perplexity has quickly become a go-to AI-powered search engine, winning fans for its summary features and fast responses.
- Scarcity of opportunities: With few AI startups raising through public markets, secondary offerings in private shares have become scarce and competitive.
- Backed by giants: High-profile investors—Nvidia, Amazon, Jeff Bezos—have taken stakes in LLM companies. Their involvement lends legitimacy and attracts other investors.
- A transformative technology: LLMs are reshaping entire industries: legal research, marketing, customer service, education, healthcare, and more. Investors see the chance to back foundational platforms that could dominate the next decade.
Their investment thesis centers on the belief that these companies are building foundational internet infrastructure, with proponents arguing that successful companies in this space could achieve valuations in the hundreds of billions.
The Skeptic’s Case: Why Some Investors Are Holding Back
Yet, others in the community remain cautious. Their hesitation boils down to three main issues:
- Talent churn and weak moats: AI talent moves fluidly between companies, and engineers can jump ship without restrictive covenants. As one member put it, “What I’m buying today may not be the same company in six months.” Users can switch between Claude, Gemini, and GPT in minutes, making defensibility more challenging.
- Sky-high valuations: Perplexity’s $20 billion valuation, on roughly $150 million in annual recurring revenue, equates to 100x+ multiples. Some investors pointed out that if the company’s free-user costs were classified differently, margins could actually swing negative. Betting at these levels requires extraordinary conviction.And yet, in context, Perplexity still looks “cheap” as OpenAI’s valuation has surged past $500 billion despite unresolved legal uncertainty about whether it’s ultimately a for-profit or a nonprofit.
- Historical parallels: One member compared today’s LLM field to the early search engine market. Back then, Yahoo, AltaVista, and Ask Jeeves looked dominant. Google was a niche startup. “Betting on one LLM today sometimes feels like buying Yahoo in 2000,” they warned.
- Unsettled infrastructure bottlenecks: GPUs, data centers, and energy supply remain the real chokepoints. If infrastructure costs remain high, model providers may struggle to generate sustainable profits.
Skeptics contend that current AI investments may mirror historical technology bubbles, with critics arguing that high valuations, weaker competitive moats, and unresolved infrastructure constraints could lead to significant value destruction for early investors.
The Alternative Play: Picks and Shovels
Faced with these risks, many investors are looking one layer down the stack: the companies that supply the infrastructure all LLMs rely on. This is the “picks and shovels” strategy, a reference to investing in suppliers during a gold rush rather than betting on who finds the biggest nugget.
Examples include, but are not limited to:
- Semiconductors
Several members brought up Nvidia ($NVDA), who is currently dominating GPU production, and Taiwan Semiconductor ($TSMC), which some members say is manufacturing at the cutting edge. Every model, including Anthropic, OpenAI, Perplexity, depends on their chips. - Data centers and deployment
Companies that solve bottlenecks in cooling, networking, and deployment cycles could stand to benefit as AI demand stretches existing capacity. Names like Nvidia (NVDA) and Arista Networks (ANET) have already seen strong tailwinds, while players in cooling and infrastructure may be next in line. - Energy providers
Uranium and nuclear power companies like Cameco (CCJ) and NexGen Energy (NXE) are drawing interest among members, given AI’s voracious power consumption. - Hybrid operators
Some members speculate that smaller, specialized AI models could shift inference from cloud hyperscalers to edge devices, changing the economics entirely. Those enabling hybrid models may become big winners, according to some members.
This approach may or may not deliver lottery-ticket outcomes, but investing in infrastructure could spread risk across the ecosystem and avoid trying to pick a single LLM champion.
How Are Investors Getting Into AI Deals?
Even for those willing to take the risk, the question remains: how do you get exposure?
- Secondary platforms like Hiive, EquityZen, and AngelList syndicates have been popular. Minimums typically range from $25K to $100K, with fees around 5–10%.
- SPVs (Special Purpose Vehicles) pool capital for a single deal, though structures vary. Some Long Angle members noted frustrations with slow allocations and shifting minimums.
- Evergreen funds and venture vehicles allow diversified access but with less concentration in headline names like Anthropic.
- Public market proxies through established tech companies like Google (Alphabet), Microsoft, and Nvidia offer immediate liquidity and indirect AI exposure, though members noted these positions provide broader technology exposure beyond just LLM development. Google, in particular, fully owns one of the leading LLMs in Gemini, and combines it with the lowest cost compute infrastructure in the industry.
These access methods typically involve trade-offs between concentration, fees, allocation sizes, and timing, with most retail investors facing structural disadvantages compared to institutional participants with direct venture relationships.
Patience vs. FOMO
For many members, this is the hardest part. Sitting out of a boom feels painful, especially when peers are announcing investments into Anthropic or Perplexity. But several investors advised caution:
- Delayed entry strategy: Some argue that market consolidation may clarify which companies have sustainable advantages, potentially offering better risk-adjusted entry points later.
- Portfolio diversification: Multiple investors emphasized spreading exposure across several companies rather than concentrating bets on individual LLM providers.
- Infrastructure focus: Others suggested that supporting infrastructure—semiconductors, data centers, power systems—might offer exposure to sector growth while reducing dependence on specific model winners.
As one member summed it up: “I wouldn’t bet on the providers. I’d bet on the infrastructure that enables them.”
Lessons From History
The most striking analogy came from the dot-com bubble. In 2000, investors piled into Yahoo and Excite. By 2005, they were irrelevant. The real winner, Google, wasn’t obvious until later.
History doesn’t repeat exactly, but it often rhymes. The AI LLM boom may produce its own Google, but Long Angle members placing big bets today know they must recognize the uncertainty.
Conclusion: Navigating the AI Investment Landscape
Current market indicators suggest LLM technology has gained significant momentum, with substantial adoption rates and continued capital investment. Many industry observers believe the technology will have lasting economic impact, though questions remain about which companies will establish sustainable competitive advantages and capture long-term value.
For investors, three paths emerge:
- Speculative bets on private LLM leaders like OpenAI, Anthropic, or Perplexity.
- Ecosystem plays in chips, data centers, public companies, and energy.
- Waiting and watching until moats become clearer and IPOs arrive.
In the words of Howard Marks, “Investing is about probabilities.” Many investors believe LLMs will reshape industries, though the odds of identifying a single long-term winner remain uncertain. For some, focusing on infrastructure rather than individual model providers feels like the more measured approach.
Frequently Asked Questions
Through secondary marketplaces, SPVs, and diversified venture or evergreen vehicles. Minimums often start around five figures and deal terms vary. Do thorough diligence on fees, transfer restrictions, and timing.
High valuations, fluid talent, evolving pricing, and limited moats. Switching costs for customers can be low. Competitive dynamics can change quickly as models leapfrog each other.
Revenue quality (ARR vs usage), gross margins after serving costs, model performance improvements over time, distribution partnerships, retention and expansion, and unit economics of inference are some of the driving factors.
Investing in the ecosystem that powers all models: semiconductors, advanced packaging, data centers, networking, and energy. This can provide exposure to model growth without picking a single winner.
No. Many investors build a barbell: a diversified basket of infrastructure and tools on one side, and selective positions in model providers or vertical apps on the other.
Many investors assume a long duration. Product cycles are fast, but business model durability takes years to prove. Size positions so you can withstand volatility and long private holding periods.
Not necessarily. Adoption and workloads continue to expand. The question is not “too late or too early,” but where in the stack and at what price.
Transfer restrictions, right of first refusal, information rights, tax treatment, fund waterfalls, and mismatch between headline valuation and security terms (common vs preferred, seniority, liquidation prefs) are some of the criteria investors use in their due diligence.
If smaller specialized models run efficiently on devices, demand could shift some workloads away from hyperscale clouds. That would change cost structures and create new winners in hardware and software.
Investors in this space often consider capping position sizes, diversifying across the stack, avoiding overpaying for growth, and revisiting theses as economics and competitive dynamics evolve.
.png)
