Who is getting rich from AI now?
In our AI infrastructure market deck, you will find everything you need to understand the market
SUMMARY
Who is getting rich from AI now? The clearest winners are the companies and people who own scarce infrastructure, powerful distribution channels or equity in the few AI businesses already scaling at exceptional speed.
AI spending reaches suppliers before it proves useful to the buyer. Chipmakers, foundries, networking vendors and power-equipment companies get paid during the buildout, while the application developer still has to show that the customer’s project saves money or earns revenue.
The same AI budget can appear as revenue several times as it moves from an end customer to an application company, a model provider, a cloud platform and an equipment supplier. That makes industry growth look larger than the original pool of customer spending and makes profit more revealing than headline revenue.
Nvidia remains the cleanest money machine because it combines huge AI exposure, unusually high operating margins and direct cash returns to shareholders. Its advantage is no longer limited to accelerators; networking, interconnects, software and complete systems deepen the amount it earns from each data-center build.
The bottleneck profit pool is widening. TSMC, Broadcom and Arista benefit from chip fabrication, custom accelerators and networking, while Vertiv and Eaton show that cooling, switchgear, backup power and electrical distribution are becoming valuable constraints too.
Hyperscalers are both winners and financiers of the boom. Microsoft, Amazon, Alphabet and Meta already earn through cloud, advertising and enterprise software, but their capital spending means reported AI growth does not immediately translate into equally strong free cash flow.
OpenAI and Anthropic are rich in revenue, funding and valuation, but outsiders still cannot see Nvidia-style economics. Their latest prices create enormous paper fortunes and limited employee liquidity, while the absence of full accounts leaves mature margins largely untested.
CoreWeave shows why commercial momentum and shareholder wealth are not the same thing. A vast backlog and high adjusted EBITDA can coexist with operating losses, heavy interest expense and depreciation that sends much of the economic value to lenders and equipment suppliers first.
Among software companies, the strongest winners control an important workflow, valuable customer data or the cost of inference. Coding has scaled fastest because usage is frequent, output is testable within minutes and the buyer can connect the bill directly to expensive developer time.
For individuals, equity matters far more than salary alone. Founders, senior researchers and early employees can capture valuation gains, while most workers see a narrower benefit through wage premiums for scarce AI skills—and those premiums do not yet amount to broad-based worker wealth.
This market map, featured in our AI infrastructure market deck, highlights top companies and startups in the AI infrastructure market
Who is getting rich from AI now?
Why is it so hard to tell who is getting rich from AI?
Today, “AI wealth” mixes cash profit, tradable shares and private valuations, so the same headline can describe three very different realities.
A company can have fast-growing revenue and still consume more cash than it produces. It can report a large adjusted profit while losing money after interest and depreciation. Its valuation can also jump by hundreds of billions of dollars without any founder, employee or investor receiving that amount in cash.
Run-rate revenue creates another source of confusion. When a company reports a $12 billion revenue run rate, it usually means its latest month multiplied by twelve. The figure shows how quickly the business is moving, though it does not prove that the company has collected $12 billion over a full year.
For this article, we give the most weight to operating profit and free cash flow. Realized share sales come next. Private valuations still matter because they can create major paper fortunes and support limited employee sales, but they provide weaker evidence than cash generated from customers.
| Type of wealth | What we look for | What it really tells us |
|---|---|---|
| Operating wealth | Operating profit, free cash flow, dividends and buybacks | The company is keeping part of the money customers spend |
| Realized personal wealth | IPO sales, acquisitions, tenders and secondary transactions | Founders, workers or investors have converted shares into cash |
| Paper wealth | Public share prices or private funding valuations | The stake is valuable at the latest transaction price |
| Commercial momentum | ARR, bookings and revenue run rates | Customers are paying, while margins and durability may remain unclear |
Where does an AI dollar go first?
An AI dollar currently reaches infrastructure suppliers earlier and more reliably than it reaches most model developers or application companies as profit.
Imagine a bank paying for an AI assistant. The application developer pays for model usage and cloud services. The model company pays for training and inference capacity. The cloud provider buys accelerators, servers, networking equipment, cooling and electrical systems.
Several companies can therefore report revenue from the same original budget. Adding every layer together would overstate the amount that the bank actually spent and the value created for the wider economy.
The order of payment also changes who takes the most risk. Nvidia, TSMC and Vertiv are paid when equipment is ordered and installed. They do not have to wait for the bank to prove that its assistant reduces costs. The application company faces a harder test because customers can cancel when the promised return fails to appear.
Cloud groups occupy a powerful middle position. Amazon, Microsoft and Google buy the hardware, rent it to model developers, sell their own AI products and sometimes own shares in the companies using their infrastructure. They can collect revenue from several parts of the same chain.
That is why the biggest verified AI fortunes are still concentrated upstream. Infrastructure companies receive cash during the buildout, while the financial outcome of many end-user projects remains unsettled.
As this chart shows, and as featured in our AI infrastructure market deck, search interest in AI infrastructure has risen sharply
Is Nvidia still the clearest AI money machine?
Nvidia is still the clearest company getting rich from AI because it combines exceptional growth with margins that no other large AI supplier currently matches.
In its latest quarter, Nvidia generated $81.6 billion of revenue, up 85% from the previous year. Data-center revenue reached $75.2 billion, accounting for roughly 92% of the total. AI infrastructure has effectively become the company.
The profit figures are even more striking. Nvidia produced $53.5 billion of operating income in one quarter, giving it an operating margin of about 66%. Its quarterly operating profit exceeded TSMC’s entire quarterly revenue and was more than twice Broadcom’s total revenue. That is an extraordinary business, even by the standards of this boom.
Nvidia is also earning more from each data center. Compute revenue reached $60.4 billion, while networking revenue reached $14.8 billion and almost tripled year over year. Customers buying huge clusters increasingly need Nvidia’s switches, interconnects, software and complete rack systems alongside its accelerators.
Shareholders are already receiving the proceeds. Nvidia returned around $20 billion through buybacks and dividends during the quarter, approved another $80 billion of repurchases and increased its dividend sharply.
Custom chips and cheaper inference will eventually put pressure on parts of this business. So far, those alternatives have expanded alongside Nvidia rather than breaking its position. No rival currently combines comparable AI revenue, growth and cash generation.
If you want more recent data on this point, please see our latest AI infrastructure market report.
Who else is getting rich from the AI chip boom?
TSMC, Broadcom and Arista are currently collecting large AI profits because they control parts of the computing stack that customers cannot replace quickly.
TSMC manufactures the most advanced chips for Nvidia, Apple, AMD and many custom-chip developers. Its latest quarterly revenue reached $40.2 billion, with a 60.3% operating margin and a 55.6% net margin. Advanced manufacturing processes produced 77% of wafer revenue. Few industrial businesses keep more than half of every sales dollar as net income.
Broadcom is winning as hyperscalers design their own accelerators. Its quarterly AI semiconductor revenue rose 143% to $10.8 billion. Across the company, adjusted EBITDA reached $15.2 billion on $22.2 billion of revenue, and management expects AI semiconductor revenue to accelerate further. Broadcom benefits whether customers buy its custom accelerators, networking components or both.
Arista occupies a smaller but highly profitable corner of the market. Its quarterly revenue grew 35% to $2.7 billion, while its operating margin remained above 42%. Thousands of accelerators are useless if they cannot exchange data fast enough, which gives high-performance networking suppliers substantial pricing power.
The common thread is hard-to-replace capacity. Technical complexity, long qualification cycles and limited supply stop customers from switching on short notice. AI demand becomes much more profitable when the seller owns one of those bottlenecks.
This chart, included in our AI infrastructure market deck, shows annual VC investment in AI infrastructure startups
Are power and cooling companies getting rich from AI too?
Yes. Power and cooling companies are now turning the data-center shortage into faster sales, larger backlogs and much stronger cash flow.
Vertiv provides cooling, power distribution and other systems used inside dense data centers. Its latest quarterly sales increased 30% to $2.65 billion. Operating profit rose 51%, adjusted operating profit climbed 64% and adjusted free cash flow jumped 147% to $653 million.
Profit grew far faster than sales. Vertiv gained from higher volumes, better pricing and the difficulty customers face when sourcing systems that can handle increasingly power-hungry AI racks. The company now expects organic sales to grow close to 30% over the full year.
Eaton is seeing the same pressure in electrical equipment. Orders in its Electrical Americas division increased 42% on a rolling twelve-month basis, while backlog grew 44%. Sales in the division rose 20% to $3.6 billion. Eaton has also bought Boyd Thermal, adding more direct exposure to data-center cooling.
These companies work with heavier equipment and lower margins than chip designers. Even so, the infrastructure windfall has clearly moved beyond GPUs. Transformers, switchgear, backup power and cooling are valuable constraints in their own right.
If you want more recent data on this point, please see our latest AI infrastructure market report.
Are Microsoft, Amazon, Google and Meta earning from AI or still funding it?
Microsoft, Amazon, Alphabet and Meta are making real money from AI today, while their enormous data-center spending delays the moment when shareholders see the full cash return.
Microsoft offers the cleanest direct disclosure. Its AI business passed a $37 billion annual revenue run rate, up 123% year over year. Azure grew 40%, and Microsoft Cloud produced $54.5 billion of quarterly revenue.
Amazon’s AWS division generated $37.6 billion of revenue and $14.2 billion of operating income. At group level, trailing free cash flow fell from $25.9 billion to $1.2 billion because purchases of property and equipment increased by $59.3 billion, primarily for AI.
Alphabet’s cloud revenue climbed 63% to $20 billion, while cloud operating income tripled to $6.6 billion. Quarterly capital expenditure reached $35.7 billion. Meta generated $56.3 billion of total revenue and $12.4 billion of free cash flow, even after spending $19.8 billion on capital expenditure.
| Company | Current AI monetization evidence | Cost of staying in the race | Our reading |
|---|---|---|---|
| Microsoft | AI run rate above $37B; Azure growth of 40% | Rapid infrastructure expansion | Strongest direct evidence of hyperscaler AI revenue |
| Amazon | AWS revenue of $37.6B and operating income of $14.2B | AI spending reduced trailing free cash flow to $1.2B | Profitable cloud business financing a much larger buildout |
| Alphabet | Cloud revenue of $20B and operating income of $6.6B | Quarterly capital expenditure of $35.7B | Cloud AI is scaling with improving segment profit |
| Meta | Advertising revenue benefits from AI recommendations and targeting | Quarterly capital expenditure of $19.8B | AI earns indirectly through an existing advertising machine |
This chart, included in our AI infrastructure market deck, shows why CoreWeave is winning in AI infrastructure
Did Amazon make more from Anthropic than from AWS?
On paper, yes: Amazon’s quarterly Anthropic gain was $2.6 billion larger than AWS operating profit, although only AWS produced recurring operating earnings.
Amazon recorded a $16.8 billion pre-tax gain from revaluing its Anthropic investment. During the same quarter, AWS generated $14.2 billion of operating income. The investment gain therefore made the larger contribution to reported profit.
No $16.8 billion payment arrived from Anthropic. The gain followed an increase in the estimated value of Amazon’s stake and could move again when the private valuation changes. AWS income came from customers paying for computing, databases, storage and other services.
Amazon still has an unusually good arrangement. It can gain when Anthropic’s valuation rises, earn cloud revenue when Claude runs through AWS, sell Trainium capacity and distribute the models through Bedrock. Its internal chip business has already passed a $20 billion annual revenue run rate.
For durable wealth, AWS operating income counts far more than a quarterly valuation gain. But the comparison shows how hyperscalers can make money from the same AI company in several different ways.
Are OpenAI and Anthropic actually rich?
OpenAI and Anthropic are currently rich in revenue and valuation, while their actual profit remains largely hidden from outsiders.
OpenAI says it now generates $2 billion of revenue per month, equivalent to about $24 billion a year at the current pace. It recently raised $122 billion of committed capital at an $852 billion post-money valuation. The same disclosure put ChatGPT above 900 million weekly users and 50 million subscribers.
Anthropic has moved even faster lately. Its revenue run rate crossed $47 billion, compared with $14 billion earlier in the year. The company then raised $65 billion at a $965 billion valuation.
Using those company figures, OpenAI’s valuation equals roughly 35 times annualized revenue, while Anthropic’s equals around 21 times. Investors are paying for the possibility that one or both companies will become dominant global platforms, well before their mature margins are known.
The size of the funding rounds also shows how expensive the race has become. OpenAI and Anthropic are committing to multiple gigawatts of computing capacity, frontier-model research and global product distribution. Revenue can grow quickly while those costs absorb a large share of it.
Both companies are commercial giants and creators of vast paper fortunes. Public evidence still falls short of showing Nvidia-style operating wealth, since neither publishes a complete quarterly income statement with comparable margins and cash flow.
If you want more recent data on this point, please see our latest AI infrastructure market report.
This chart, included in our AI infrastructure market deck, shows annual funding in AI infrastructure startups
Is CoreWeave actually making money from the AI cloud boom?
CoreWeave currently moves enormous amounts of AI money, while common shareholders are still waiting for dependable profit.
Quarterly revenue more than doubled to $2.08 billion, and revenue backlog reached $99.4 billion. CoreWeave has also passed one gigawatt of active power and secured more than 3.5 gigawatts of contracted capacity. Customer demand is clearly large.
Adjusted EBITDA reached $1.16 billion, giving the company a headline margin of 56%. Further down the accounts, CoreWeave reported a $144 million operating loss, $536 million of net interest expense and a $740 million net loss. Adjusted operating income was only $21 million.
The gap comes from the economics of an asset-heavy business. CoreWeave must finance data centers and accelerators, then depreciate that equipment as newer chips arrive. Interest and depreciation consume a large part of the amount left after day-to-day operating costs.
A $99.4 billion backlog offers strong revenue visibility, yet it cannot tell us whether every contract will earn an attractive return after financing, electricity and hardware replacement. Today, lenders and equipment suppliers have clearer claims on CoreWeave’s cash than its common shareholders do.
Which AI software companies have become real businesses?
A small group of AI software companies has already become commercially substantial, especially where the product controls important data or a costly daily workflow.
Databricks has passed a $5.4 billion revenue run rate, growing more than 65% year over year. Its AI products account for $1.4 billion, net revenue retention remains above 140%, and the company produced positive free cash flow over the previous twelve months. A more recent funding term sheet valued it at $188 billion.
Palantir provides the strongest public proof. Quarterly revenue increased 85% to $1.63 billion. Operating income reached $754 million, and adjusted free cash flow reached $925 million. The company is converting AI demand into cash rather than relying solely on a higher share price.
The younger application companies are smaller but growing quickly. ElevenLabs increased ARR from $350 million to more than $500 million in four months. Glean reached $300 million of ARR, three times its level fifteen months earlier. Harvey reports 98% gross revenue retention, 167% net dollar retention and 77% seat utilization in legal and professional services.
| Company | Best current commercial evidence | Financial quality | Our judgment |
|---|---|---|---|
| Databricks | $5.4B revenue run rate; growth above 65% | Positive free cash flow and retention above 140% | Proven enterprise platform expanding through AI |
| Palantir | $1.63B quarterly revenue; growth of 85% | $754M operating income and $925M adjusted free cash flow | Strongest profitable public AI application company |
| ElevenLabs | More than $500M ARR | Full profit figures remain private | Voice AI has become a large enterprise category |
| Glean | $300M ARR, tripling in 15 months | Full profit figures remain private | Enterprise context is supporting repeatable growth |
| Harvey | 98% gross retention and 167% net dollar retention | Limited public financial disclosure | Strong product use, with less visibility on overall economics |
This chart, included in our AI infrastructure market deck, compares the main business model options for AI cloud infrastructure providers
Why is coding AI producing winners so quickly?
AI coding is currently the fastest application market because developers use it every day, the budget is easy to justify and the output can be tested immediately.
Anthropic says Claude Code has passed a $2.5 billion revenue run rate after more than doubling since the beginning of the year. Weekly active users also doubled, business subscriptions quadrupled and enterprise customers now produce more than half of Claude Code revenue.
That single product already generates several times the disclosed ARR of leading voice, enterprise-search and legal AI companies. Coding reached this scale faster because the customer usually understands the product without a long corporate transformation program.
Developers are expensive, and even a modest improvement in output can cover a subscription or usage bill. Their work is already digital. Generated code can be run, tested and corrected within minutes, giving both the user and the model a rapid feedback loop.
Usage is also unusually frequent. A developer may call a coding agent hundreds of times during a project. That produces recurring consumption and creates a natural route from an individual account to a team contract and then to an enterprise agreement.
Competition will remain intense because model providers, independent coding companies and large software platforms all want this market. The demand itself is no longer in doubt. Coding has become the first AI-native application category to reach multi-billion-dollar scale through ordinary work usage.
Can AI applications keep SaaS-like margins?
AI applications can reach strong software margins, but they need control over the workflow, the customer’s data or the cost of inference.
Traditional SaaS companies can serve another user at very little additional cost. An AI application may pay for tokens every time somebody generates code, searches documents or runs an agent. Heavy usage helps revenue while also increasing the supplier bill.
The strongest businesses are finding ways around that problem. Palantir reported a gross margin near 87% because customers pay for a broad operational and data platform. Databricks has reached positive free cash flow while allowing companies to manage data, models and governance through one system.
Glean offers a more direct example of cost control. In a company benchmark, its context system used 30% fewer tokens than off-the-shelf alternatives while receiving higher user preference scores. Better retrieval and context can reduce the number of expensive model calls needed to complete a task.
Other applications can protect margins by routing simple work to cheaper models, caching common answers, using smaller internal models or charging according to the financial outcome delivered. Legal, coding and customer-service tools can also command higher prices because they replace expensive professional time.
A generic interface built on somebody else’s model has little room to defend its price. The application companies most likely to keep the money will own the customer relationship and something important underneath it.
If you want more recent data on this point, please see our latest AI infrastructure market report.
This chart, featured in our AI infrastructure market deck, shows the share of revenue generated by each customer segment in the AI infrastructure market
Are consultants getting rich from corporate AI confusion?
Consultants are earning real AI money now, although Accenture’s latest results resemble a healthy services business more than an explosive new profit pool.
Accenture generated $18.7 billion of quarterly revenue, up 3% in local currency. Operating margin reached 17%, and free cash flow was $3.6 billion. Management reported more large-scale AI transformation programs and 104 client bookings worth at least $100 million during the year to date, up 13%.
Corporate AI projects often require much more than choosing a model. Companies need to clean data, connect old software, control access, redesign workflows, train employees and satisfy regulators. Consultants can bill across all of those areas.
They also receive payment during the implementation period. The client may spend another year determining whether the project improves profit, while the consulting revenue has already been recognized.
The economics remain tied to people and project delivery. Accenture’s latest local-currency growth was modest compared with the growth of leading chip, cloud and software businesses. Its 17% operating margin also sits far below the margins earned by Nvidia, TSMC, Palantir and several other bottleneck owners.
Consulting firms are clear beneficiaries of AI complexity. Their upside remains capped by headcount-heavy delivery unless they turn more of their work into repeatable software or managed platforms.
Who is rich on paper, and who has actually cashed out?
Founders and early employees at leading AI companies have become extremely wealthy on paper, while tender offers and secondary sales are converting only part of those holdings into cash.
At the latest transaction prices, OpenAI and Anthropic are each valued close to $1 trillion. Databricks has signed a funding term sheet at $188 billion. Those prices make early stakes extraordinarily valuable, even when the owners cannot freely sell them.
Private companies increasingly use tender offers to give workers some liquidity. ElevenLabs authorized a $100 million employee secondary transaction at a $6.6 billion valuation before its ARR and valuation rose further. Databricks has also said that part of its financing will support employee liquidity.
These programs usually limit how many shares each person can sell. Founders and early executives benefit most because they own larger stakes, while later employees may hold fewer shares with higher exercise prices.
Venture funds face the same distinction. A new financing round can raise the reported value of a portfolio overnight. Fund investors receive spendable money only when shares are sold through a secondary transaction, acquisition or public listing.
The AI boom has already created genuine personal fortunes. Still, the combined cash withdrawn by founders, workers and venture investors remains much smaller than the wealth implied by the latest private valuations.
This chart, included in our AI infrastructure market deck, shows how GPU cloud infrastructure technology has evolved over time
Are companies using AI getting richer than companies selling it?
The next great AI fortunes may come from ordinary companies using AI, but today’s evidence still favors suppliers and digitally mature incumbents.
PwC’s latest AI Jobs Barometer examined more than one billion job advertisements and company performance across multiple countries. Companies most able to use AI recorded headcount growth of 52%, compared with 36% among the least AI-exposed companies. Their wage growth was also higher, at 24% versus 17%.
The difference was much larger among the strongest performers. The top fifth of companies in highly AI-exposed sectors achieved average labor-productivity growth of 163% from their 2018 baseline, nearly five times the rate for the broader group of AI-exposed companies.
Those figures still need a careful reading. Companies that deploy AI well may already have better management, cleaner data, stronger software systems and more money to invest. The study shows a widening gap rather than proving that AI alone caused every percentage point.
Meta, Alphabet and Amazon illustrate how incumbents can turn AI into revenue through existing businesses. Better recommendations, advertising, search and logistics affect ordinary financial lines, which makes the AI contribution difficult to separate.
The shift toward users has started, but suppliers still capture the easiest cash. Every adopter must buy infrastructure or software before knowing how much value the project will create.
Are ordinary workers getting richer from AI?
AI is making a narrow group of workers richer now, especially people who combine scarce technical skills with valuable industry judgment.
PwC found that jobs requesting specific AI skills offered an average wage premium of 62%, up from 57% in the previous edition of its study. AI-related job postings grew 69%, while the overall job market grew 9%.
The premium does not mean that adding “prompt engineering” to a résumé raises anyone’s salary by 62%. Many of these roles combine machine learning, software engineering, product responsibility and specialist knowledge. Employers are paying for rare combinations of skills.
Entry-level work is changing too. AI-exposed entry roles requiring traditionally senior abilities such as judgment and leadership grew 35% from 2019, while other entry-level roles declined 10%. Employers appear to expect junior workers to handle higher-level decisions sooner.
Equity creates the biggest upside. Founders, senior researchers and early startup employees can benefit from the full increase in a company’s valuation. A salaried employee without shares receives only the wage negotiated with the employer.
For most workers, higher productivity can lead to several outcomes: better pay, more demanding targets, reduced hiring or fewer hours. Current evidence supports a large reward for scarce AI expertise, while broad worker wealth remains much less visible.
In our AI infrastructure market deck, we identify pain points entrepreneurs should prioritize
So who is getting rich from AI now?
AI is already making people rich, and the biggest verified fortunes currently belong to infrastructure owners, hyperscalers and a small group of enterprise software companies.
Nvidia sits at the top because it combines huge AI revenue with exceptional operating margins and direct cash returns to shareholders. TSMC, Broadcom and Arista control other scarce parts of the computing stack. Vertiv and Eaton show that the windfall now reaches cooling and electrical equipment as well.
Microsoft, Amazon, Alphabet and Meta form the next group. Their existing customers, clouds, advertising systems and balance sheets let them earn from AI today while financing the next wave of infrastructure. Their spending is enormous, yet they already own profitable businesses underneath it.
Palantir and Databricks provide the strongest evidence among enterprise software companies. Coding has become the fastest new application category, while voice, enterprise search and legal AI are developing into sizable businesses. The winners tend to own valuable data, a daily workflow or a direct relationship with corporate buyers.
OpenAI and Anthropic have created extraordinary commercial momentum and some of the largest private fortunes ever recorded. Their valuations tell us how much investors expect them to become worth. The absence of full public accounts leaves their current profit much harder to judge.
Consultants are being paid, selected employees are selling shares and people with scarce AI skills receive major wage premiums. CoreWeave and many application startups show the other side of the boom: fast revenue can coexist with heavy financing costs or uncertain margins.
The answer is fairly blunt. The people getting richest from AI now generally own a bottleneck, a distribution channel or equity in one of the few companies scaling unusually fast. Intelligence attracts the attention, but ownership decides who keeps the money.
If you want more recent data on this point, please see our latest AI infrastructure market report.
OUR METHODOLOGY
We separated AI wealth into four categories: operating wealth, realized personal wealth, paper wealth and commercial momentum. We gave the most weight to operating profit and free cash flow, followed by dividends, buybacks and completed share sales. Revenue run rates and private valuations were used to show scale and expectations, not treated as proof of durable profit.
We followed the AI dollar through the value chain to distinguish companies that are paid during the infrastructure buildout from those that must still prove an end customer earns a return. This also prevents the same customer budget from being counted as fresh economic value at every layer of chips, cloud, models and applications.
For public companies, we prioritized the latest earnings releases, investor presentations and regulatory filings. For private companies, we used company funding announcements, disclosed run rates, tender offers and secondary transactions. Adjusted metrics were included where useful, but we read them alongside operating income, interest, depreciation and free cash flow.
We used cross-company comparisons only when they clarified the economics: Nvidia’s profit scale, CoreWeave’s gap between adjusted EBITDA and net income, hyperscaler AI revenue against capital expenditure, and application revenue against retention or cash generation. The aim was to separate a large business from a business that is already keeping a large share of the money.
Key sources include Nvidia quarterly results, TSMC investor relations, Broadcom quarterly results, Arista investor relations, Vertiv investor relations, Eaton investor relations, Microsoft earnings, Amazon investor relations, Alphabet investor relations, Meta investor relations, CoreWeave investor relations, Palantir investor relations, company newsrooms for OpenAI, Anthropic, Databricks, ElevenLabs, Glean and Harvey, plus PwC’s AI Jobs Barometer and Accenture investor relations.
This chart, included in our AI infrastructure market deck, shows the share of revenue by region across Europe, Asia, North America, Africa, and South America in the AI infrastructure market
Related blog posts
- Who is making money in digital health now?
- Is the AI infrastructure boom a bubble?
- AI cloud: which startup is ahead?
