The Next Trillion-Dollar Company: Why It Will Likely Be AI

The next trillion-dollar company is more likely to be an AI company than anything else, and the queue is already forming. As of 31 October 2026, SpaceX (which now owns the xAI lab) has listed at around $2.1 trillion, Anthropic has made its IPO filing public, and OpenAI is reportedly raising money at a $1.25 trillion valuation. That is a prediction about where the money is flowing, not a promise that these valuations will hold. I am a student developer, not an analyst, so this post sticks to what I could verify from filings and news reports, dates everything that changes, and keeps my own view for the end. I also treat the energy that powers AI as one layer of the story, because it limits how fast the rest can grow.

Who is in the trillion-dollar club today, and how did they get there?

The club is bigger than most people assume, and its exact membership changes whenever markets move. A 24/7 Wall St. list published on 27 May 2026 counted 14 companies: Nvidia, Apple, Microsoft, Alphabet, Amazon, Broadcom, TSMC, Meta, Tesla, Saudi Aramco, Berkshire Hathaway, Samsung, SK Hynix and Micron. Other lists differ at the edges, and SpaceX joined after that list was written. The same article notes that the club had exactly one member in 2018, and that was Apple. According to CMC Markets' list of milestone dates (published December 2025, updated October 2026), the crossings went like this: Apple on 2 August 2018, Amazon on 2 September 2018, Microsoft on 25 April 2019, Alphabet on 16 January 2020, Meta on 28 June 2021, Tesla on 25 October 2021 and Nvidia on 13 June 2023. The pattern is easy to see. Apple sold a device and then an ecosystem around it, Amazon and Microsoft built cloud businesses, and Alphabet and Meta sold advertising at enormous scale. Nvidia is the clearest AI-era entry. For the quarter that ended on 26 July 2026, Nvidia reported revenue of $96.2 billion, up 106% from a year earlier, with $89.0 billion coming from its data center segment. Its GAAP gross margin was 75.0% and its GAAP net income was $59.7 billion. It guided to $108 billion of revenue for the next quarter and said that outlook assumes no China data center compute revenue.

What does a trillion-dollar valuation actually require?

A company's market value is its profit multiplied by the multiple investors are willing to pay for that profit. So a trillion-dollar valuation needs a large profit, a growth story that justifies the multiple, or both. The numbers below are hypothetical and only show the shape of the arithmetic.

Text

Hypothetical example, not a forecast. Figures in billions of dollars.

Target market value:                 1,000
Assumed price-to-earnings multiple:  25
Net profit needed:                   1,000 / 25 = 40

Revenue needed for that profit:
  at 10% net margin:  40 / 0.10 = 400
  at 20% net margin:  40 / 0.20 = 200
  at 30% net margin:  40 / 0.30 = about 133

Years to grow from 20 of revenue to 200 (a 10x):
  at 60% growth a year:  about 4.9 years
  at 40% growth a year:  about 6.8 years
  at 25% growth a year:  about 10.3 years

Margin matters as much as size, because a business that keeps 30 cents of each dollar needs a third less revenue than one that keeps 20. Growth rates also fade, so a company that doubles every year at 20 billion rarely keeps doing that at 200 billion. You can also change the multiple. At 50 times earnings the same company would need only 20 billion of profit, but a high multiple is a bet on future growth, and it shrinks quickly when growth slows. Nvidia shows what the top of the range looks like. Its latest quarter earned about 62 cents of net profit per dollar of revenue (59.7 divided by 96.2), which is an unusually high figure for a company of that size.

What are the layers of the AI economy?

It helps to split AI into layers, because the economics differ a lot from one to the next. Chips and hardware sit at the bottom. This is Nvidia for the processors, Broadcom and others for custom chips and networking, TSMC for manufacturing, and memory makers such as SK Hynix, Micron and Samsung. Revenue here is already large and profitable, which is why several of these names are in the club. Cloud and data centres come next. Microsoft, Alphabet, Amazon and Meta are spending at a scale that has few precedents. According to CNBC's report on 28 July 2026, Alphabet had raised the top of its 2026 capital spending range to $205 billion, Microsoft had projected $190 billion including finance leases, Amazon had guided to $200 billion, and Meta was expected near $139 billion. The same report noted that long-term debt had risen at Alphabet and Amazon and that analysts expected free cash flow to turn negative at some of them. Foundation models form the third layer. These are the labs that train the large models: OpenAI, Anthropic, Google and others. Bloomberg reported, via TechCrunch, that Anthropic's annualized revenue passed $65 billion at the end of July 2026, up from $9 billion at the end of 2025. A run rate extrapolates one recent period, so it is not the same as audited yearly revenue. CNBC reported in March 2026 that OpenAI earned $13.1 billion in 2025 and was still burning cash. Applications are the top layer: the products built on those models, from coding assistants to customer support agents to tools inside ordinary business software. I wrote about what this means for existing software firms in is SaaS dying.

Can the power grid and natural gas keep up with AI?

Every layer above depends on electricity, and electricity is the slowest layer to build. The International Energy Agency's April 2025 Energy and AI report estimated that data centres used about 415 TWh in 2024, around 1.5% of global electricity, and projected that to more than double to roughly 945 TWh by 2030. The United States accounted for 45% of the 2024 total and for nearly half of the projected growth in US electricity demand. Natural gas is a big part of the near-term answer in the US. The IEA expected gas-fired generation to grow by about 175 TWh to serve data centres, with renewables covering about half of the global growth and nuclear adding a similar amount to gas. Berkeley Lab's December 2024 report for the US Department of Energy found data centres used about 4.4% of US electricity in 2023 and could reach between 6.7% and 12% by 2028. The IEA's follow-up, published on 16 April 2026, carries a press release headline saying data centre electricity use surged in 2025 even as bottlenecks tightened. The binding constraint is equipment and wiring, more than fuel. The 2025 report said transmission lines can take four to eight years to build in advanced economies, that wait times for transformers and cables have doubled in three years, and that around 20% of planned data centre projects could face delays. In July 2026, Utility Dive reported that GE Vernova's gas turbine backlog had reached 116 GW and that it was taking reservations for 2031 deliveries, with about 20% of its gas turbine customers being data centre companies. My inference from this is that a company can raise billions in a week, but it cannot get a power connection in a week. That caps how quickly the whole system can grow. I would guess the energy layer is less likely to produce a new trillion-dollar company than the layers above it, since much of it is regulated utilities and equipment makers with long lead times.

Which company could be the next trillion-dollar company?

Start with the ones that may already have crossed or are very close, because the question changed during 2026. SpaceX listed on Nasdaq on 12 June 2026 at $135 a share and closed its first day at a value of about $2.1 trillion, after raising $75 billion. CNBC presents it mainly as a rocket and satellite business, with an AI unit that came from the acquisition of xAI and that leases computing capacity to other companies. So it is only partly an AI company. OpenAI closed a round of $122 billion at a post-money valuation of $852 billion on 31 March 2026, according to CNBC. On 31 October 2026, CNBC relayed Financial Times reporting that it is seeking up to $20 billion more at a $1.25 trillion post-money valuation, and that the round had not closed. Anthropic raised $65.2 billion in a Series H, which media reports put at about $965 billion after the money went in. According to Invezz's reading of the filing, it made its S-1 public on 26 October 2026 without a price range, and the reports of a $1.5 trillion to $2 trillion IPO value come from IFR and Reuters, according to Invezz, and have not been confirmed by Anthropic. Private valuations come from negotiations, and a public market can reprice them on the first day. Where do incumbents have the advantage? Nvidia, Broadcom and TSMC already earn real profits from the hardware layer, and the four cloud giants have the data centres, the customers and the cash. If AI keeps growing, the companies already in the club are the likeliest to stay there. A newcomer to the trillion-dollar list is therefore most likely to be a model lab, because it sits in the layer with the fastest reported revenue growth and the most to gain from scale. An application company is harder to picture at that size. It depends on a model supplier that can change prices, and it competes with every other team using the same models. That is my inference, and a few companies may prove it wrong.

Why is it so hard to build a trillion-dollar AI company?

Capital intensity is the first problem. Training and running frontier models needs chips, data centres and power before the revenue arrives, and the cloud giants are now borrowing and spending down cash to do it. A lab that is not one of them has to raise money almost continuously, which is what the OpenAI and Anthropic rounds above are. Compute cost is the second. OpenAI reported $2 billion of revenue per month in March 2026 and was still not profitable, according to CNBC. If each answer costs real money to produce, growth in usage can widen losses before it narrows them. Competition is the third, and it works through price. The IMF's October 2026 Global Financial Stability Report, as reported by The Star, said there is growing evidence that AI capabilities can be delivered at lower investment levels. If that holds, the models become cheaper for everyone and the margin per answer shrinks. Regulation is the fourth, and Nvidia's guidance, which assumes no China data center compute revenue, shows how much a single market can matter. Talent is the fifth. Researchers who can train frontier systems are scarce and expensive, though I do not have a verified number to give you for that.

Is AI a bubble?

Here is the strongest version of the bear case. Spending is running far ahead of proven profit, and parts of the financing are circular: companies that sell chips and cloud capacity also invest in the labs that buy them, which can make demand look larger than it is. The IMF named an AI-driven correction among its main near-term risks in October 2026, and it said weaker or less durable AI investment and profitability could trigger it. It did not say a crash is coming. Google's chief executive Sundar Pichai told the BBC in November 2025 that there are elements of irrationality in the boom and that no company would be immune if it burst. Here is the strongest version of the bull case. The money at the centre is real. Nvidia earned $59.7 billion in one quarter, which is not a promise or a forecast. If you took that quarter's profit and multiplied it by four, then at the $5 trillion level Nvidia first passed on 29 October 2025 (as widely reported at the time), you would get a multiple of about 21. That is my own calculation from two numbers on different dates, and it ignores that profits may fall or rise. The demand side shows up in other places too: GE Vernova is booking gas turbines years ahead, and Anthropic's reported run rate rose from $9 billion to $65 billion in seven months. Both cases can be true for different companies, since a buildout can be useful and still leave some of its financiers with losses.

What do railroads, the dot-com crash and mobile teach us?

Railroads are the closest case for infrastructure. During Britain's Railway Mania of the 1840s, Parliament authorised 263 railway companies in 1846 alone, and roughly a third of the authorised lines were never built. Even so, about 6,220 miles of line was built from the 1844 to 1846 authorisations, in a country whose network is now roughly 11,000 miles. The crash hurt investors, and the infrastructure stayed. The dot-com era shows how long a right idea at a wrong price can take to recover. Cisco's share price hit a record close of $80.06 in March 2000, and according to Sherwood News it matched that level again only on 10 December 2025. Its market value was still about $250 billion below the peak of more than $550 billion, because it had bought back so many shares. The internet won, and the leading seller of its equipment still took 25 years to get its share price back. Mobile is the more hopeful case. The iPhone arrived in 2007, and Apple became the first trillion-dollar US company in 2018, about eleven years later. The biggest winner of mobile was a platform owner rather than a network equipment company. None of these maps cleanly onto AI. Nvidia reports very large profits today, which is a real difference from the weakest dot-com companies, while OpenAI, for one, was still burning cash as of March 2026. Treat each case as a reminder of one risk and one hope, and use none of them as a forecast.

What can builders and developers do about it?

You do not need to pick a winner to make good decisions. These questions are a practical way to judge any AI plan, including your own:

Text

1. Which layer am I in: chips, cloud, model, or application?
2. Who sets my main cost, and can they raise the price?
3. What does one request cost me, and what do I charge for it?
4. Can I switch model providers in a day, or is it a rewrite?
5. What do I own that a competitor on the same model cannot copy?

If you build applications, you are almost certainly in the top layer and renting the layers below, so your edge is your data, your workflow and your relationship with customers. Keep the model call behind one small module so a provider or price change does not touch the rest of the code, and log the cost of each feature from the start. Any AI feature I add to a project like GoCart, whose AI features are still in development, would sit in this layer, which means I would be a customer of the model labs and not a competitor. If you are choosing a stack for an early product, I covered the trade-offs in MVP tech stack for startups. For the chip layer in more depth, see Nvidia's future, and for the wider benefits and risks, the future of AI.

My take

I am a final-year Computer Science student (more about me on the about page), so treat this as an informed outsider's view and nothing more. The evidence supports the claim that a trillion-dollar AI-native company is plausible, and the private valuations already reported in October 2026 put OpenAI and Anthropic near that line. I would expect the first clear case to be a model lab or something close to one, because that is where reported revenue is growing fastest. What remains uncertain is whether profits arrive on a schedule that supports the valuations. Capital spending is huge, some of the financing is circular, and the IMF is already pointing at a correction as a risk. If models get cheaper and competition squeezes margins, the value could end up with the chip makers and the power equipment suppliers, or be spread across users of AI rather than concentrated in a few firms. Three things would change my mind: a lab reporting steady profit alongside rising revenue, falling capital spending without falling demand, or power constraints that stop growth for years. I could be wrong on all of this, and the dates in this post will go stale quickly. This is not financial advice.

Want to talk about this?

I am a full stack developer in Rahim Yar Khan, Pakistan, a final-year Computer Science student, and I work in web development at Hadi Dev Lab. If you want to compare notes on building products on top of AI models, you can message me on WhatsApp or by email. You can also see my work. Cover photo from Unsplash.

The Next Trillion-Dollar Company: Why It Will Likely Be AI

Who is in the trillion-dollar club today, and how did they get there?

What does a trillion-dollar valuation actually require?

What are the layers of the AI economy?

Can the power grid and natural gas keep up with AI?

Which company could be the next trillion-dollar company?

Why is it so hard to build a trillion-dollar AI company?

Is AI a bubble?

What do railroads, the dot-com crash and mobile teach us?

What can builders and developers do about it?

My take

Want to talk about this?

Recent posts