Nvidia 投資 OpenAI 對 AI 主題的影響

就在 Broadcom 和 Oracle 宣布從 OpenAI 獲得長期且重大的 AI GPU 與數據中心訂單後，不少人擔心 Sam Altman 是否會減少向 Nvidia（以及 AMD）購買 GPU。僅僅幾週後，Nvidia 宣布將向 OpenAI 投資 1000 億美元，用於建設一個 10 兆瓦的 AI 數據中心項目。

這顯然超出了 Nvidia 早前的業務預測；同時也疊加了 OpenAI 與 Oracle 及 Broadcom 的其他 AI 基礎設施計劃。OpenAI 這些龐大的 AI 項目規模令人震撼，也強烈暗示未來對計算能力的需求將會極為龐大。

Nvidia 並非「自己買自己的晶片」

部分人對此交易持懷疑態度，認為 Nvidia 是透過向 OpenAI 注資，間接促成對自家晶片的採購，質疑這是否屬於「真實銷售」。關鍵在於理解這是兩筆獨立的交易：OpenAI 是一項極具吸引力的私募股權投資標的，而 Nvidia 作為 AI 領域的巨頭，深知並重視其主要客戶，因此希望參與投資。

此外，OpenAI 的計算需求極為龐大，可能只有 Nvidia 能夠滿足。建設一座 10 兆瓦的 AI 數據中心，預估需要 400 至 500 萬顆 GPU，而能供應此規模的廠商極為有限，尤其考慮到 OpenAI 還有其他長期的 AI 基礎設施計劃。儘管該項目將分年建設，但這 400 至 500 萬顆 GPU 的需求，幾乎等同於 Nvidia 一整年的銷售與生產能力。

粗略估算，依照目前市場價格，10GW 數據中心的建設成本約為 5,500 億美元，其中約 3,500 億美元將用於採購 Nvidia 的 GPU 及其他網路與系統設備。Nvidia 的 1,000 億美元投資遠不足以支付整體費用，且此次交易並非「自行購買晶片」。儘管此項投資可能有助於 Nvidia 獲得「優先供應商」資格，但面對如此龐大的合約，OpenAI 的選擇空間本就非常有限。

OpenAI 將籌集並賺取現金以支付龐大合約

就如 Oracle 的 3,000 億美元數據中心合約一樣，有人質疑 OpenAI 是否具備足夠現金埋單，畢竟其目前現金儲備有限。儘管這類龐大的長期合約在規模與時間上具有彈性，供應商仍會對買方的支付能力進行盡職調查。

對於像 OpenAI 這樣成功的公司而言，其在中長期內預期將籌集並產生大量現金，因此供應商如 Nvidia、Oracle、Broadcom 和台積電對其履行財務義務的能力充滿信心。

在即將或正在進行的融資輪中，OpenAI 可能籌集 300 至 500 億美元，估值約達 5,000 億美元。若 Nvidia 的 1,000 億美元投資屬於此輪融資，OpenAI 有可能籌得更多資金，因為其現有主要支持者（如微軟與頂尖私募基金）可能追加投資，以避免股權被稀釋。

若這筆 1,000 億美元投資分散於數年內進行，也意味著其他投資者在未來融資輪中將投入更多資金，以支撐 OpenAI 的長期發展與基礎設施建設。

OpenAI 或將成為美國首個 1 兆美元 IPO

Sam Altman 可能計劃成為美國首個 1 兆美元 IPO，這意味著本輪 5,000 億美元的融資可能是 OpenAI 上市前的倒數第二輪或最後一輪融資。憑藉 OpenAI 的聲譽和成長潛力，有機會成功做到10,000億美元總市值的IPO，因為全球大多數主要基金、捐贈基金和退休基金都會在其上市前後競相追捧。

正因如此，執行長 Altman 才能夠在中期將成長置於利潤之上，因為他能夠籌集足夠的資金來支持擴張，而無需依賴營運現金流。只要 OpenAI 在人工智慧領域的領導地位仍然對資金具有吸引力，它就能擁有充足的資本來繼續投資和發展其人工智慧帝國。身為人工智慧產業的領導者，Altman 先生深知人工智慧的潛力，也深知建構強大基礎設施以在這場重要的人工智慧革命中保持領先地位的必要性。

AI主題和Nvidia的更多上漲空間

這也顯示了AI硬體和基礎設施的巨大成長潛力，因為OpenAI、微軟、Google、亞馬遜、和Meta Platform等巨頭不斷擴大其AI資本支出。許多投資人低估了未來5-10年AI運算需求的大幅成長，希望OpenAI、軟銀、微軟、Meta Platform等公司近期發布的公告能幫助他們更好地掌握AI的未來並進行相應的投資。

作為業界領先的硬體供應商，Nvidia的每股盈利預計將從去年的3.0美元成長50%至今年的4.50美元，並在明年（2027年1月31日）再成長33%至6.0美元。以目前183美元的股價計算，這相當於2025年收益的40.7倍市盈率和2026年收益的30.5倍市盈率。對於一家領先的、最重要的人工智慧電腦硬體寡頭供應商來說，儘管股價處於歷史高位，但其估值仍然頗具吸引力，甚至可以說是便宜。未來12個月，其股價容易達到210美元，相當於2026年每股盈利的35倍。在明年銷售額成長率維持在30%以上的多頭情況下，其本益比可能達到40倍，即240美元。

Nvidia-OpenAI US$100 billion investment deal’s implication to the AI theme

Only a couple weeks after Broadcom & Oracle announcing securing major long-term AI GPU & datacentre orders from OpenAI that sparked concern if Sam Altman will be buying less GPU from Nvidia (& AMD), this ‘monumental’ deal of Nvidia investing US$100 billion into OpenAI to build a 10-gigawatt AI datacentre project. Apparently, this is in addition to Nvidia’s earlier business projection; also, on top of OpenAI’s other AI infrastructure plans with Oracle & Broadcom. The size of these mega AI projects from OpenAI is overwhelming and a strong indication of how big is the future demand of computing capacity.

Nvidia stock price jumps 3.4% to new high US$183 and boost TSMC 3.9% to new record US$275, Oracle gains 4.5% to US$322 tonight. ASML & Applied Materials keep their 3% & 5% rise from analyst upgrade before this Nvidia-OpenAI news. NASDAQ expand its intraday gain by 143 points to new high of 22,790 while S&P500 is up 32 to 6,689 and may challenge 6,700 for the first time in history.

Nvidia is NOT buying its own chips

Some people find this deal sceptical as it seems like Nvidia injecting money into OpenAI to buy its own chips and question if this is ‘real sales.’ It is key to understand the two parts are independent transactions; OpenAI is a highly sought after PE investment assets that Nvidia would like to invest as the AI giant understand and appreciates its major client very well. Also, OpenAI has gigantic compute demand that maybe only Nvidia can meet. Building a 10-gigawatt datacentre requires 4-5 million GPUs which limited providers can supply especially as OpenAI also has other long-term AI infrastructure plans. Obviously, this will be built & spread over a few years but this 4-5 million GPUs project is equivalent for the full-year sales & production capacity of Nvidia of current year.

Roughly speaking, 10GW datacentre costs US$550 billion to build as per current prices, and US$350 billion of Nvidia’s GPU & other networking & system products. US$100 billion investment is not enough to pay the bill and Nvidia is not ‘buying its own chips’ through this mega transaction. While the investment plan may help facilitate the ‘Preferred Supplier’ status for Nvidia, there may not be many options on supplier for such huge contract.

OpenAI would raise & generate cash to pay the bill

Of course, like the Oracle’s US$300 billion datacentre contract, some doubt if OpenAI will have the cash to pay the bill as it does not have that much cash now. Although such huge long-term contracts are flexible in size and timing but suppliers would perform due diligence on buyer’s payment capability. For a successful firm like OpenAI which will raise & generate huge amount of cash in the mid-to-long term, suppliers like Nvidia, Oracle, Broadcom, & TSMC are confident about their ability to fulfil their financial obligations.

In upcoming or current round of capital raising, OpenAI may raise US$30-50 billion for a valuation of US$500 billion. If Nvidia’s US$100 billion investment is for this round, then OpenAI would raise more as most of its existing backers such as Microsoft & top PE funds would top up to avoid dilution. If this $100 billion is spread over a few years, that should mean other investors would invest substantial capital in future rounds too.

OpenAI may target to become US first US$1 trillion IPO

Sam Altman maybe planning to become the first US$1 trillion IPO listing in United States and that means the current US$500 billion round could be the second last or final round before public listing. With OpenAI reputation and growth potential, even a US$1 trillion total market cap upon IPO could be a successful one as most major funds, endowment, & pension funds around the world would be chasing its stocks before and after public listing.

That is why CEO Altman can afford to prioritize growth over profit in the mid-run as he can raise sufficient capital to fund expansion and does not need to rely on operation cash flow. As long as its AI leadership remains appealing to funds, it could have abundant capital to keep investing and growing its AI empire. As the AI industry leader, Mr Altman understands the potential of AI and the need to build a strong infrastructure to keep its lead in this important AI revolution.

More upside for AI theme & Nvidia

This also demonstrates the huge growth potential of AI hardware & infrastructure as the industry gurus including OpenAI, Microsoft, Google, Amazon, Meta Platform, & Nvidia keep expanding their AI capital expenditure. Many investors underestimated the big jump in AI compute demand in the next 5-10 years and hope these recent announcements from OpenAI, Softbank, Microsoft, Meta Platform, & etc. have helped them get a better grasp of the AI future and invest accordingly.

As the industry’s top hardware providers, Nvidia is expected to grow its EPS by 50% from US$3.0 last year to US$4.50 this year and then another 33% to US$6.0 next year (31 Jan 27). At current stock price of US$183, that translate to 40.7x PE on CY2025 earnings & 30.5x PE on CY 2026 profit. For a leader and an oligopoly provider of most important AI computer hardware, it looks attractive or even cheap despite at record high. In the next 12 months, it can easily reach US$210, or 35x 2026 EPS. In a bull case scenario of sales growth staying above 30% next year, it could fetch 40x PE or US$240.