NVIDIA overtakes Microsoft to become world's most valuable company
NVIDIA ended June 18, 2024 as the world’s most valuable listed company, ahead of Microsoft and Apple. The milestone combined actual revenue with future expectations; separating the two is essential to reading market capitalisation without mistaking it for cash.
NVIDIA ended the June 18, 2024 trading session as the world’s most valuable listed company. Its shares closed at $135.58 and its market capitalisation reached roughly $3.334 trillion, ahead of Microsoft at $3.317 trillion and Apple at $3.286 trillion. The ranking was correct when the bell rang, but it did not mean NVIDIA held that amount in a bank account or earned more than its rivals. It was the price at which the market valued all its shares at that moment.
The distinction matters because a stock-market ranking blends present and future. Results explained an extraordinary part of the rise: quarterly revenue had more than tripled year over year. The rest was a collective expectation about how many AI accelerators NVIDIA would sell, how long its advantage would last and how much profit each sale would produce. Reading the figure requires separating the accounting fact from the thesis embedded in the price.
How $3.334 trillion is constructed
Market capitalisation is calculated by multiplying the price of one share by the number of shares outstanding. The Associated Press closing measurement recorded $3.334 trillion for NVIDIA, $3.317 trillion for Microsoft and $3.286 trillion for Apple. Reuters, using LSEG market data, put NVIDIA at $3.335 trillion and confirmed the same order. The small difference is rounding, not a contradiction.
The marginal price—the quotation for the last share traded—is applied to every share to obtain the total. That does not imply all holders could sell simultaneously at $135.58: a wave of supply would move the price. Market capitalisation is useful and comparable, but it is not a cash balance, liquidation appraisal or guaranteed acquisition cost.
Nor does it directly measure operating size. A company can produce more revenue and be worth less if markets expect slower growth or lower margins. Another can lead the ranking with fewer sales because investors anticipate much larger future profits. “Most valuable” here means the highest market capitalisation among listed companies, not the largest workforce, revenue, assets or social importance.
The stock split created no value
NVIDIA had completed a ten-for-one stock split after trading closed on June 7. Every holder received nine additional shares for each one owned, and split-adjusted trading began on June 10. NVIDIA’s official split explanation is explicit: neither the total value nor each shareholder’s proportional interest changed.
A pizza cut into ten times as many slices contains no more food. Where there had been one share worth roughly $1,200, there would be ten worth roughly $120, before ordinary market movements. The unit price falls and the share count rises in the same proportion; their product, market capitalisation, remains unchanged at the moment of adjustment.
The split can make a whole share more accessible to employees and smaller investors, NVIDIA’s stated purpose. It does not by itself explain the 3.5% price rise on June 18 or why the total value passed Microsoft’s. Confusing accessibility of one share with creation of wealth is one of the most common errors in reading market news.
The business really did change scale
NVIDIA’s results released on May 22 provide the observable foundation. In the first quarter of fiscal 2025, ended April 28, the company reported $26.044 billion in revenue—262% more than a year earlier and 18% more than the previous quarter. Data Center contributed $22.6 billion, up 427% year over year.
That was about 87% of quarterly revenue. The transformation was not merely narrative: Data Center had displaced gaming as the dominant engine. NVIDIA’s 10-Q filed with the SEC attributed growth to higher shipments of the Hopper platform for training and inference of large language models, recommenders and generative applications, as well as expanding InfiniBand networking sales.
The same filing adds two qualifications a record-setting headline can miss. Large cloud providers represented roughly 45% of Data Center revenue. Two direct customers accounted for 13% and 11% of total quarterly revenue, while two indirect customers each exceeded 10%. Powerful demand can also be concentrated: if a handful of buyers delay investment, the effect reaches a supplier quickly.
Margins strengthened the case. Quarterly gross margin reached 78.4%, up from 64.6% a year earlier, driven by Data Center growth. GAAP net income was $14.881 billion. These figures help explain why the market did not price NVIDIA like an ordinary hardware manufacturer: exceptional expansion was arriving with exceptional profitability.
Why a GPU became infrastructure
A GPU performs many operations in parallel. That design, born for graphics, fits the matrix multiplications that dominate neural-network training and inference. Yet NVIDIA’s position did not depend on one chip. It included CUDA, libraries, interconnects, networking and complete systems; switching suppliers could require changes to software and operations, not merely replacing a card.
Hopper drove current results while Blackwell supported part of the expectation. NVIDIA had introduced Blackwell on March 18 as its next Data Center platform, combining GPUs, NVLink interconnects, networking and software. The company claimed reductions of up to 25 times in cost and energy for specified inference workloads against the preceding generation. “Up to” and “specified” are essential: this was a vendor claim about particular configurations, not a guaranteed saving for every model.
The product transition offered opportunity and risk. If customers wanted the new generation, NVIDIA could renew sales; if manufacturing, integration or supply slipped, a gap could emerge. The 10-Q said H100 supply was improving while H200 remained constrained and warned that adding suppliers and capacity commitments increased complexity. The market was rewarding future execution that still had to happen.
A valuation is a hypothesis with a price
To analyse a milestone like this, it helps to draw three columns. The first contains closed facts: revenue, profit, margins, business mix and customer concentration. The second contains operating indicators: manufacturing capacity, supply, adoption, software and competition. The third contains expectations: cloud capital spending, the profitability of AI services and demand for the next architecture.
Market capitalisation reacts when any column changes. A brilliant quarter may not lift a share if it was already priced in; a merely adequate one can drive gains if it beats a low expectation. A higher price therefore does not prove that a technology works socially or that chip buyers will recover their investment. It shows that marginal buyers accepted a higher price at that moment.
The company must also be separated from the industry. Demand for AI computing could grow while NVIDIA’s share fell to AMD accelerators or hyperscalers’ custom chips. NVIDIA could remain strong even if some generative products were unprofitable, provided other computing workloads sustained spending. The causal chain must be tested, not treated as part of a label such as “AI fever.”
The transferable skill is to read market capitalisation as the product of shares and price, then divide that price into demonstrated results and pending expectations. NVIDIA truly was number one by market value on June 18, and its accounts explained a substantial part of the ascent. But the record was neither accumulated cash nor a guarantee. It was a market hypothesis about sales, margins, competition and execution, expressed in a figure that could change in the next session.
This article was produced with artificial intelligence under human editorial oversight.
