Speakers in this discussion argue AI infrastructure spending faces a timing gap, since months can pass between when computing hardware ships and when it starts earning revenue, leaving a large amount of capital tied up. As an example, a buyer of $10 million in Nvidia chips may need six months to get them running while that money sits on its balance sheet. They also point to a wider gap between total AI spending and the revenue it generates, and note that some reports suggest the size of returns may depend on technological innovation. Easing supply chain bottlenecks, they say, would help companies bring computing capacity online faster.
Summary written by Upside AI from the original article. Read the source for full details.
Read the original on Yahoo Finance ↗
Keywords
NVDAAI investmentartificial intelligencedata centerssupply chain
Open a keyword to see other news on the same topic.
Related tickers and likely impact
NVDA
• NeutralNvidia
Relevance: medium
| Market | US stocks |
|---|---|
| Sector | Information Technology |
| Industry | Semiconductors |
| Market cap | $5.2T |
| Close on 10/20/2026 | 216.48 |
| Change that session | -0.28% |
Impact is scored by AI from the headline and may be wrong. It is not investment advice.
Related news
- Tech companies need more data centers . So why are nearly half of all planned projects delayed or canceled ?
Spending on AI infrastructure is projected at about $700 billion through global hyperscale spending, yet building U.S. data centers is hitting bottlenecks in industrial supply chains and grid capacity rather than chip shortages. Industry research cited in the article finds that 30%–50% of the 16 GW of U.S. data center capacity scheduled to come online faces severe delays or cancellation. Causes include lengthy permitting, a shortage of roughly 439,000 skilled construction workers, electrical equipment deficits, and grid interconnection queues of four to seven years in Northern Virginia, Phoenix and Dallas. Oracle and OpenAI scrapped plans to expand the Crusoe Energy-developed campus in Abilene, Texas, from 1.2 GW to 2.0 GW after talks over financing and utility timelines stalled.
- Nvidia announces plug and play Vera Rubin racks that can be shipped on a truck
Nvidia announced the DSX Pod AI Factory, which puts seven Vera Rubin AI racks into a unit roughly the size of a standard shipping container and weighing about 28,000 kilograms. It is designed to be trucked on normal roads, with cabling and liquid cooling assembled at the factory so that on-site work is limited to placing, connecting and validating the units. Nvidia says this cuts the time for a new AI data center to start serving tokens from six months to 14 days. No price was disclosed, though analysts say a single Vera Rubin rack can cost over $7 million. The pod is a reference design built by partners including Foxconn, Gigabyte, Lenovo, HPE, Quanta and Pegatron, with deliveries expected in early 2027.
- Nvidia's Jensen Huang wants to build AI factories in weeks, not months
Nvidia CEO Jensen Huang, speaking at Nvidia GTC Berlin, said the new DGX Pod is designed to build AI factories faster and more cheaply, potentially cutting deployment from months to weeks. He said a gigawatt of data center costs about $50 billion, and the 4 to 6 months from shipping components to operation could be shortened by 3 to 4 months if more manufacturing happens in the supply chain, reducing idle cash. Huang argued this would make financing easier because lenders would have more certainty about costs and when revenue starts, which matters amid market volatility over neocloud companies. He said demand is very strong, citing SpaceX, CoreWeave, Nebius and Nscale, and that the large capital tied up before revenue is the key challenge to solve.
- An AI bust could come from places investors arent watching , an economist says
Concerns about an AI bubble are growing, and TS Lombard's Dario Perkins warns that investors may be overlooking other risks. Debt among smaller cloud providers, model developers and suppliers could reach $2 trillion by the end of 2027. Perkins does not see this as a systemic risk, but warns that earnings disappointments and canceled orders could ripple through the ecosystem, and that a chip glut could follow over-ordering. Much of the industry's revenue growth still comes from spending within the AI ecosystem rather than from businesses and consumers outside it. He also flags political backlash, noting that ordinary people dislike AI and nearby datacenters, and warns that one high-profile accident could turn AI toxic.
- Tech companies need more data centers. So why are nearly half of all planned projects delayed or canceled?
- How Investors Are Reacting To NVIDIA (NVDA) AI Partnership Wave
Adobe, Delta Electronics, Navitas, London Stock Exchange Group and other partners have deepened collaborations with NVIDIA, tying RTX Spark chips, Nemotron models and CUDA-X software into AI content creation, semiconductor workflows and 800 VDC data center power designs. The article argues these partnerships embed NVIDIA's AI platform more deeply in customers' daily workflows, though they do not clearly change the core catalyst of converting backlog into shipped AI factory capacity. The main risks remain export controls on China and large customers building in-house accelerators. Under the investment narrative cited, forecast revenue reaches US$995.9 billion and earnings US$540.6 billion by 2029, based on assumed annual revenue growth of 48.7%.
Related markets
Latest news
- Amazon (AMZN): Can Custom AI Silicon Outpace Capex Commitments?
Amazon opened its Trainium AI chips to every PyTorch developer, letting engineers run existing code on its own hardware without rewriting it. The chips business is selling at an annual pace above $25 billion, and AWS sales grew 37% in the second quarter, up from 17% a year earlier, while CEO Andy Jassy said the AI business has also passed a $25 billion annual run rate. Anthropic and OpenAI have made multi-year commitments to Trainium. However, AI-related spending pushed free cash flow to an outflow of $7.6 billion over the past 12 months, versus an inflow of $18.2 billion a year earlier, and long-term debt rose to $128.9 billion from $65.6 billion at the end of last year. Second-quarter operating income rose 43% to $27.5 billion, but the article says the stock now depends on AWS demand lasting long enough for cash flow to recover.
- Anthropic Has Locked In Over $600 Billion of Computing, More Than Intel Is Worth. It Still Wants More and That’s Big News for AMD.
Ahead of its IPO, Anthropic has committed more than $600 billion to computing, more than Intel's $484.7 billion market value, according to leaked reports of its S-1 filing. The article cites three different totals: $518 billion of signed commitments (Reuters, late September), above $600 billion (Crypto Briefing), and a possible $750 billion of extra computing spending over several years (Wall Street Journal). Anthropic reportedly brought in $16.3 billion of revenue in the first half of 2026 but posted an $80.7 billion net loss, mostly from accounting charges. AMD said Anthropic will deploy up to 2 gigawatts of its MI450 GPUs starting in the first half of 2027, and AMD's second-quarter revenue rose 50.1% to $11.54 billion, with Data Center revenue up 107% to $6.72 billion. The article says the risks center on Anthropic's hard-to-cancel commitments, its losses and its dependence on an IPO, and notes AMD trades at a P/E of 201.
- AI bottlenecks 'keep shifting': Nvidia CEO Jensen Huang explains
Nvidia CEO Jensen Huang, speaking with Deirdre Bosa at Nvidia GTC Berlin, said AI bottlenecks keep shifting, moving from wafers to CoWoS, HBM memory, silicon photonics and connectors, and now to power. He said power is often available, but land, power shells, generator delivery, grid connections and construction limits create the problem, and permitting rules differ across countries and states. Huang argued Nvidia is resilient because it supports every AI model, cloud provider and computer company across every country, so it can ship chips wherever a bottleneck is resolved. He expected these physical bottlenecks to persist for a couple more years as AI factories ramp up.
- Nvidia (NVDA) Is Trading Like Growth Will Stop. The Numbers Say Otherwise.
A hotter-than-expected inflation report pushed Nvidia down more than 3% in one session, as investors worried about interest rate hikes that would raise borrowing costs for companies building AI data centers. Fiscal second-quarter revenue still reached $96.2 billion, up 106%, with data center sales up 117%, and the company guided fiscal third-quarter revenue to $108.0 billion, plus or minus 2%, excluding data center compute sales to China. Gross margin was 75.0% in the quarter. The article's main question is whether Nvidia's customers can keep paying for chips, pointing to CoreWeave's sharp drop after its bond yield chart spread online and the doubts it raised about its heavy debt. Nvidia has announced financing partnerships with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR aimed at bringing more than $500 billion of outside capital into AI infrastructure over time, though definitive agreements are still needed.
- Bitcoin price tags $84.6K as US stocks head higher on SpaceX Q3 earnings beat
Bitcoin rose nearly 2% to $84,600, its first time at that level since Oct. 7, as US stocks gained after SpaceX reported third-quarter revenue of $14.6 billion, up 161% year over year and beating expectations by nearly 15%. SpaceX also posted net income of $4.4 billion, and analysts at Truist Securities said its results helped calm fears over corporate spending on AI buildouts. The tech-heavy Nasdaq Composite rose 0.7% early in the session before giving back some gains, while South Korea's KOSPI finished 3.29% higher at 5,735 points. Rekt Capital said Bitcoin needed a daily close above $83,400 to shift focus higher, and that a rejection there could send price below $82,500.
- Bybit will look for big broker partnerships to take on Robinhood in the U.S.
Bybit CEO Ben Zhou said the exchange plans to enter the U.S. through a partnership with an established local firm, such as Interactive Brokers, rather than pursue state-by-state licenses that could take two to three years. He noted Bybit, which has over 80 million registered users in more than 180 countries and regions, has a clean U.S. record and would wait for the Clarity Act or a clearer federal regulator before moving rather than rushing in. Zhou described the U.S. as competitive and saturated, and pointed to the U.K. as a relatively untapped market with a new crypto regime and no dominant global exchange, though entry costs are very high. Bybit also plans to launch an everything app offering derivatives and tokenized stocks in Europe as early as the first quarter of next year.
Get the briefing on Telegram or Zalo
Join the Upside Telegram channel or Zalo group for the Vietnam stocks and gold briefing at 8:00 and 19:00 (Vietnam time) on trading days, plus a weekly wrap on Saturday morning.