DeepSeek Faces Computing Limits as AI Demand Strains Memory Supply

AI infrastructure demand is driving sharply different company updates across chips, memory and networking. Cisco reported revenue of $18.5 billion, up 24% and its fastest growth in more than 15 years, and expects $9 billion in AI sales to cloud companies this fiscal year. It raised its revenue forecast to $74.8 billion–$75.8 billion, though its shares fell about 6.5% on Nov. 13 as margins narrowed amid higher AI hardware and memory-chip costs. Cisco generated $4.4 billion in operating cash and returned $3.7 billion to shareholders last quarter.

Applied Materials posted record quarterly revenue of $10.41 billion, up 53%, and adjusted earnings per share of $4.16, up 92%. Its stock nevertheless dropped 2.7% after results and remained about 50% below its 52-week high. CEO Gary Dickerson expects chipmaking equipment spending to rise more than 30% in 2027; cloud companies are expected to increase data center spending by more than 50% next year. Applied expects packaging revenue to grow about 90% this calendar year and plans to double systems output by 2028. A separate report says it added more than 10 factory projects to its tracking list for a third consecutive quarter.

Memory makers face strong AI demand but also production and market risks. SK Hynix leads high-bandwidth memory (HBM), having begun mass-producing HBM3 in 2022; Micron started HBM3E production in 2024, and both companies moved to HBM4 in early 2026. HBM production limits capacity for conventional DRAM and NAND, and analysts expect growth to slow. KB Securities estimates China accounts for about 29% of global memory demand and says its AI infrastructure spending could cushion weaker U.S. demand. U.S. export controls restrict China’s access to advanced accelerators and HBM, prompting wider use of domestic chips such as Huawei Ascend. The report cites DeepSeek’s computing-resource shortages as evidence of unmet demand.

Networking suppliers also point to AI data-center demand. Ciena, whose shares have fallen almost 40% in three months, expects fiscal fourth-quarter 2026 revenue of $1.75 billion, up 30% year over year, and a 20% non-GAAP operating margin versus 13.2% a year earlier. It forecasts more than $10 billion in backlog at fiscal year-end and fiscal 2027 revenue of $8.3 billion–$8.4 billion. Ciena expects its addressable market to reach $52 billion by 2029, growing at a 24% annual rate, and plans to report results on Dec. 10.

The semiconductor sector has also sold off: the PHLX Semiconductor Sector index is down 25% since early October. Yet global chip revenue reached $159.7 billion in August 2026, up 144% year over year, and the Semiconductor Industry Association expects full-year revenue to exceed $1.6 trillion. TSMC, which manufactures chips for Nvidia, AMD and Broadcom, holds 73% of the pure-play foundry market; its revenue rose 43% in the first 10 months of 2026. One report cites a $580 median 12-month analyst price target for Applied Materials, implying 66% upside, while Morgan Stanley projects the wafer and fab equipment market will reach $149 billion in 2026 and $191 billion in 2027.

Data-center operators CoreWeave and Nebius are using customer prepayments to help finance expansion as investors worry about debt. CoreWeave says customers paid upfront on 70% of its 2026 deals and expects to fund most of its 2027 spending through internal cash and customer payments. Nebius says payments on new contracts cover more than 100% of the cost of bringing contracted capacity online; it expects $17 billion–$20 billion in customer payments next year against estimated capital spending of $40 billion–$45 billion. Both companies have raised prices as demand remains strong and capacity tight. Separately, OpenAI CEO Sam Altman warned that chip suppliers may be counting on the same limited set of buyers, including OpenAI and Anthropic, creating a risk of short-term supply-demand mismatch. He said data-center demand remains solid and computing needs should grow over time.

Several other developments span AI investment and deployment. Nvidia is in advanced talks to buy two-year-old diffusion-model startup Inception for at least $500 million, according to people familiar with the discussions; Google and Microsoft have also invested in the company. Alphabet may be better positioned than smaller rivals if tighter AI rules or data-center construction limits emerge, and Google Cloud revenue rose 103% to $30.7 billion in the third quarter, supported by its Tensor Processing Unit systems. The supplied CoreWeave stock-pick articles offer no detailed comparison with IREN or Nebius, but describe the Motley Fool Stock Advisor’s shift from AI “Act 1” research toward “Act 2” global rollout and cite its past Netflix and Nvidia recommendations.

Funding and product launches add to the activity. On Nov. 18, 25 startups raised a combined $2.2 billion; Shield AI raised $1 billion for autonomy systems and intelligent aircraft, while Palo Alto Networks acquired Applied Compute for $4.5 billion. On Nov. 19, six startups announced $366.2 million in funding, led by Beijing-based MegaRobo Technologies’ $200 million round and Mistral AI’s $110 million. DriveCentric and FordDirect expanded access to DriveCentric’s sales and service platform and four AI agents for Ford and Lincoln dealers. Databricks says its AI Gateway Observability Insights tool identified $499,000 in token waste tied to seven MCP server bugs and 12,000 engineer waiting hours; fixes took about an hour.

AI’s effects are also reaching consumers and markets. Standard Life found that half of UK adults who used AI for pension information say it influenced their savings choices, including 58% of Gen Z and Millennials; it cautions users to verify answers because they can be wrong or outdated. Man Group says Japanese AI hardware suppliers’ share of the TOPIX fell from about 18% at June’s peak to around 11.5% during November’s decline, while some optical-cable and automotive-chip companies traded at 10–14 times expected earnings. An article on Anthropic’s stock-market debut raises questions about AI valuations and whether the sector is a bubble, while arguing that AI could reshape industries; it says time will tell.

Key Takeaways

  • Cisco’s revenue rose 24% to $18.5 billion; it expects $9 billion in AI sales to cloud companies this fiscal year, despite margin pressure from AI hardware and memory costs.
  • Applied Materials recorded $10.41 billion in quarterly revenue, up 53%, and expects chip-equipment spending to rise more than 30% in 2027; it plans to double systems output by 2028.
  • SK Hynix leads HBM, while Micron and SK Hynix have both moved to HBM4; HBM production constrains conventional memory capacity and analysts expect growth to slow.
  • KB Securities estimates China represents about 29% of global memory demand and says domestic AI investment, including use of Huawei Ascend chips, could support memory demand; the report cites DeepSeek’s computing limits.
  • Ciena forecasts fiscal Q4 2026 revenue of $1.75 billion and more than $10 billion in year-end backlog; it reports results Dec. 10.
  • CoreWeave says customers prepaid on 70% of its 2026 deals. Nebius says new-contract prepayments cover more than 100% of capacity-delivery costs and expects $17 billion–$20 billion in customer payments next year.
  • OpenAI CEO Sam Altman warned that chip suppliers may rely on the same limited buyers, including OpenAI and Anthropic, risking short-term mismatches between supply and demand.
  • Nvidia is in advanced talks to acquire diffusion-model startup Inception for at least $500 million; Google and Microsoft have also invested in it.
  • Google Cloud revenue grew 103% to $30.7 billion in the third quarter. TSMC, which makes chips for Nvidia and AMD, holds 73% of the pure-play foundry market.
  • Startup funding included Shield AI’s $1 billion round on Nov. 18 and MegaRobo Technologies’ $200 million and Mistral AI’s $110 million rounds on Nov. 19. Databricks says its observability tool found $499,000 in token waste.

Cisco’s AI Networking Growth Makes It a Calmer Stock Pick

Cisco reported revenue of $18.5 billion, up 24% and its fastest growth in more than 15 years. It expects $9 billion in AI sales to cloud companies this fiscal year, more than double last year, and raised its full-year revenue forecast to $74.8 billion to $75.8 billion. The stock fell about 6.5% on Nov. 13 as profit margins narrowed, partly because Cisco sold more AI hardware and paid more for memory chips. Cisco also generated $4.4 billion in cash from operations last quarter and returned $3.7 billion to shareholders through dividends and stock buybacks.

SK Hynix and Micron Race to Supply AI Memory Chips

SK Hynix and Micron have benefited from rising demand for high-bandwidth memory, or HBM, used in AI data centers, but both stocks have fallen over the past two months. SK Hynix was the first company to mass-produce HBM3 chips in 2022 and now leads the HBM market; Micron began mass-producing HBM3E chips in 2024 and moved to HBM4 in early 2026. HBM production is reducing their capacity to make conventional DRAM and NAND chips. Analysts expect growth at both companies to slow over the next few years, though the article favors SK Hynix because it is larger and growing faster.

Applied Materials Sees Strong Demand for Chipmaking Equipment

Applied Materials reported record quarterly revenue of $10.41 billion on Nov. 12, up 53% from a year earlier, while adjusted earnings per share rose 92% to $4.16. The stock fell 2.7% the next day and was about 50% below its 52-week high. CEO Gary Dickerson expects chipmaking equipment spending to rise more than 30% in 2027, while cloud companies are expected to increase data center spending by more than 50% next year. Applied expects its packaging business to grow about 90% this calendar year and plans to double systems output by 2028.

Applied Materials Reports Record Sales as Factory Demand Grows

Applied Materials posted record quarterly revenue of $10.41 billion on Nov. 12, a 53% increase from a year earlier. Adjusted earnings per share climbed 92% to $4.16, but the stock fell 2.7% the next day and remained about 50% below its 52-week high. CEO Gary Dickerson expects spending on chipmaking equipment to rise more than 30% in 2027, and management added more than 10 factory projects to its tracking list for the third quarter in a row. The company expects its packaging business to grow about 90% this calendar year and plans to double systems output by 2028.

Ciena’s AI Networking Demand Could Boost Its December Results

Ciena plans to report quarterly results on Dec. 10 as demand grows for optical networking equipment used in AI data centers. The company expects fourth-quarter fiscal 2026 revenue of $1.75 billion, up 30% from a year earlier, and a non-GAAP operating margin of 20%, compared with 13.2% last year. Ciena expects its addressable market to grow at a 24% annual rate from 2026 to 2029, reaching $52 billion, and forecasts revenue growth of 30% per year over that period. It expects to end fiscal 2026 with more than $10 billion in backlog and forecasts fiscal 2027 revenue of $8.3 billion to $8.4 billion.

Ciena’s sharp sell-off may set up a rebound in December

Ciena shares have fallen almost 40% in three months, but the company is seeing strong demand for optical networking equipment used in AI data centers. Ciena expects fourth-quarter fiscal 2026 revenue of $1.75 billion, up 30% from a year earlier, and more than $10 billion in backlog by the fiscal year’s end. It expects revenue to rise more than 30% in fiscal 2027 and operating margins to grow over the next several years. Ciena will report quarterly results on Dec. 10, and the article says its stock could benefit from the report.

CoreWeave, IREN and Nebius are compared as AI stock picks

The article raises the question of whether investors should buy CoreWeave, but the supplied text does not include a comparison of CoreWeave, IREN and Nebius. It describes the Motley Fool Stock Advisor’s view that AI is moving from an early research stage, called “Act 1,” toward a global rollout, or “Act 2.” It cites past recommendations of Netflix on Dec. 17, 2004, and Nvidia on April 15, 2005. A $1,000 investment in those recommendations would have grown to $321,850 and $955,232, respectively, according to the article.

Alphabet may benefit from tighter AI rules and data center limits

After Democrats won November’s midterm elections, investors worried that new leaders might tighten AI rules and slow data center construction. The article says Alphabet could benefit because stricter rules may make it harder for smaller AI companies to compete, while Alphabet has legal, financial and lobbying resources. Alphabet also has existing data centers and capital to handle construction delays, which could make its permitted facilities more valuable. In the third quarter, Google Cloud revenue rose 103% to $30.7 billion, and the company’s Tensor Processing Unit systems helped drive sales growth.

TSMC and Applied Materials stand out after chip stock sell-off

The PHLX Semiconductor Sector index has fallen 25% since early October, while global semiconductor revenue reached $159.7 billion in August 2026, up 144% from a year earlier. Taiwan Semiconductor Manufacturing makes chips for companies including Nvidia, Advanced Micro Devices and Broadcom, and holds 73% of the global pure-play foundry market. Its revenue grew 43% in the first 10 months of 2026, and the article says its earnings could rise further as it plans higher prices and forecasts strong AI accelerator growth. Applied Materials reported record fourth-quarter fiscal 2026 revenue of $10.41 billion, up 53%, and analysts set a 12-month median price target of $580, implying 66% upside.

CoreWeave, IREN and Nebius are compared as AI stock picks

The article asks whether investors should buy CoreWeave, but the supplied text does not provide details comparing the three companies. It says the Motley Fool Stock Advisor team views the AI industry as moving from an early research phase, “Act 1,” to a global rollout, “Act 2.” The article points to earlier recommendations of Netflix on Dec. 17, 2004, and Nvidia on April 15, 2005. It says a $1,000 investment in those recommendations would have grown to $321,850 and $955,232, respectively.

CoreWeave and Nebius use customer payments to fund AI expansion

CoreWeave and Nebius Group rent out data center capacity for artificial intelligence, but their shares have fallen amid concerns about debt. In 2026, CoreWeave received upfront payments on 70% of its customer deals and expects to fund most of its 2027 spending through its own cash and customer payments. Nebius says new contract payments cover more than 100% of the cost to bring promised capacity online, and it expects $17 billion to $20 billion in customer payments next year. Both companies have raised prices as demand for AI data center space grows and supply remains tight.

CoreWeave and Nebius strengthen funding for AI data centers

CoreWeave and Nebius Group rent data center capacity to customers running artificial intelligence workloads. Their shares have fallen as investors worry about the debt needed to build more facilities, but both companies are using customer prepayments to help pay for expansion. CoreWeave has collected upfront payments on 70% of its deals in 2026 and expects to fund most of its 2027 capital spending internally and through customer payments. Nebius says prepayments cover more than 100% of the cost of bringing contracted capacity online and expects $17 billion to $20 billion in customer payments next year. Both companies are charging higher prices as demand remains strong and data center capacity is limited.

CoreWeave and Nebius raise prices as AI data center demand grows

CoreWeave and Nebius Group provide dedicated data center capacity for artificial intelligence customers. Investor concerns about the debt needed to build their data centers have weighed on both stocks, so the companies are turning to customer prepayments to help fund expansion. CoreWeave says customers paid in advance on 70% of its deals in 2026, while Nebius says payments on new contracts cover more than 100% of the cost of bringing that capacity online. Nebius expects $17 billion to $20 billion in customer prepayments next year, while both companies are charging more for capacity as demand outpaces supply.

CoreWeave and Nebius rely on prepayments to limit data center debt

CoreWeave and Nebius Group rent data center capacity for artificial intelligence, and both stocks have fallen amid concerns about debt-funded expansion. CoreWeave says it has collected customer prepayments on 70% of deals in 2026, with customers sometimes paying the full contract value upfront. Nebius says prepayments on new contracts cover more than 100% of the capital needed to bring that capacity online, compared with 50% to 60% in the second quarter. The company expects $17 billion to $20 billion in customer prepayments next year, against estimated capital spending of $40 billion to $45 billion. Both providers have raised prices as demand for AI data center capacity remains strong and supply is tight.

Chip stock sell-off puts TSMC and Applied Materials in focus

The PHLX Semiconductor Sector index has fallen 25% since early October, as investors booked profits in Nvidia shares and worries about inflation and Federal Reserve policy weighed on the market. The Semiconductor Industry Association reported that global chip revenue rose 144% year over year in August 2026 to $159.7 billion, and it expects revenue to exceed $1.6 trillion for the full year. The article highlights Taiwan Semiconductor Manufacturing, or TSMC, and Applied Materials as companies that could benefit from industry growth. TSMC makes chips for Nvidia, Advanced Micro Devices, and Broadcom, and has 73% of the global pure-play foundry market. Applied Materials supplies equipment used to manufacture chips, while Morgan Stanley forecasts the wafer and fab equipment market will grow to $149 billion in 2026 and $191 billion in 2027.

SK Hynix and Micron Ride HBM Demand, but SK Hynix Holds an Edge

SK Hynix and Micron have both gained from rising demand for high-bandwidth memory, or HBM, used in AI data center GPUs. SK Hynix was the first company to mass-produce HBM3 chips in 2022 and still leads the global HBM market. Micron skipped HBM3 but began producing HBM3E in 2024, and both companies moved to HBM4 in early 2026. Their growth has been strong, but analysts expect it to slow, while inflation, interest rates, and possible limits on data center construction could weigh on AI spending and their stock values.

Shield AI Leads 25 Startup Funding Rounds Raising $2.2 Billion

Twenty-five startups announced funding rounds on November 18, 2026, raising a combined $2.2 billion. Shield AI raised $1 billion for its AI-powered autonomy systems and intelligent aircraft. Other large rounds included $280 million for a London-based business spend management platform, $126 million for Japan’s Helical Fusion, and $124 million for space computing company Ramon.Space. The roundup also reported that Palo Alto Networks acquired Applied Compute for $4.5 billion.

MegaRobo Technologies Leads Six Startup Funding Rounds

Six startups announced funding on November 19, 2026, raising a total of $366.2 million. MegaRobo Technologies, a Beijing-based biotech company using AI and robotics for drug research, raised $200 million. Mistral AI of Paris raised $110 million to develop and provide AI models, assistants, and agents. Other disclosed rounds included $48.5 million for a factory-direct sourcing platform, $4.7 million for a private equity due diligence platform, and $3 million for technology that helps brands improve visibility in AI assistants. Algoryx also raised an undisclosed amount.

KB Securities Says China AI Spending Could Support Memory Demand

KB Securities said China’s investment in its own AI infrastructure could help support global memory demand, even if U.S. spending slows. Analyst Lee Chang-min estimated that China accounts for about 29% of global memory demand and named Samsung Electronics and SK Hynix as the firm’s top picks. U.S. export controls limit China’s access to advanced accelerators and high-bandwidth memory, so China is using domestic chips such as Huawei Ascend and working to improve and scale its systems. Lee said uncertainty over U.S. AI rules could continue to affect memory stocks through year-end, but may ease after the new Congress convenes on January 3.

China’s AI Spending Could Help Offset a U.S. Memory Demand Slowdown

KB Securities said China’s AI infrastructure investment could cushion the global memory market if U.S. demand weakens. The firm estimates China makes up about 29% of global memory demand, and analyst Lee Chang-min said China’s growing needs could limit any overall decline. U.S. export controls restrict China’s access to advanced accelerators and high-bandwidth memory, so the country is expanding use of domestic chips such as Huawei Ascend and building larger systems. The report pointed to DeepSeek’s limits on usage because of a shortage of computing resources as evidence of unmet demand. It said U.S. rules are more likely to impose conditions, such as requiring data center operators to pay for electricity, than to broadly restrict AI investment, while market volatility could last through year-end as investors await clearer policies.

DriveCentric brings sales, service and AI agents to Ford and Lincoln dealers

DriveCentric announced on Nov. 19, 2026, that Ford and Lincoln dealers can now use its full platform through an expanded partnership with FordDirect. The platform includes Sales Engagement Hub, Service Engagement Hub and four AI agents: Prospect Agent, Nurture Agent, Sales Agent and Service Agent. The hubs bring customer conversations across sales and service together, while the agents can reach out and follow up around the clock using shared data. Dealers can purchase the solutions through FordDirect, and DriveCentric and FordDirect will hold a webinar on Dec. 9 at 2 p.m. ET.

Sam Altman warns AI chip suppliers may be counting on the same buyers

At Semafor’s Nov. 16 Silicon Valley & the World event, OpenAI CEO Sam Altman warned that many chip suppliers expect to sell large volumes to the same buyers, including OpenAI and Anthropic. He said those buyers and data centers have limited demand, raising the risk that supply and demand may not match in the short term. Altman also said demand for data centers remains solid and that computing demand should grow over the long term. His comments came as AI stocks fell.

Japan’s AI hardware shares fall as investors look beyond banks

Man Group says the decline in Japanese technology shares may create opportunities in AI hardware, industrial and automotive companies. Japanese banks have risen since inflation returned and the Bank of Japan began raising interest rates, but they now trade at nearly twice their book value and make up more than 40% of the main Japanese large-cap value index. Man Group says higher interest rates, near 1.5%, may leave banks with less room to repeat recent gains. AI hardware suppliers’ share of the TOPIX index fell from about 18% at June’s peak to around 11.5% during November’s market decline. The article says some optical cable and automotive chip makers traded at 10 to 14 times expected earnings, while some memory chip makers lost as much as 70% of their value from June highs.

Databricks tool identifies $1.2 million in AI agent waste

Databricks says its AI Gateway Observability Insights tool can analyze live Unity Gateway traffic and rank token waste by cost. The company says seven MCP server bugs led to $499,000 in wasted tokens and 12,000 engineer waiting hours, and fixing them took about an hour. Insights compares each workstream’s share of sessions with its share of tokens, and teams can add custom questions to reports or ask Genie to identify waste and draft bug reports. It can also assess task complexity to help teams route routine sessions to cheaper models. The tool stores traces in a governed Delta table in Unity Catalog and can create reports covering the past day, seven days or 30 days.

Half of UK adults using AI for pension information say it affects savings choices

Research from Standard Life found that 50% of UK adults who have used AI for pension information say it influenced their retirement savings decisions. The share was 58% among Gen Z and Millennials, compared with 27% of Gen X and 25% of Baby Boomers. In the past year, 29% of Gen Z used AI for pension information, compared with 16% of UK adults overall. Common questions concerned how pensions work and how much money people may need in retirement, both at 34%. Standard Life says AI can help people understand pensions, but users should check answers with trusted sources and providers because information may be wrong or out of date, and scams can be convincing.

Anthropic’s Stock Market Debut Raises Questions About AI Investing

Anthropic, the company behind Claude, is described as the first generative AI company to enter the stock market. Its debut has renewed questions about whether AI is an investment bubble, as some people point to fast rising valuations. The article compares AI with tulip mania but argues that AI represents a new economy with the potential to change industries and create opportunities. It says only time will show whether AI is a bubble or a new frontier.

Nvidia in Advanced Talks to Buy AI Startup Inception for at Least $500 Million

Nvidia is in advanced talks to acquire Inception, a two year old startup that develops diffusion based AI models. The proposed deal is worth at least $500 million, according to three people familiar with the discussions. Inception’s technology is designed to work faster and more efficiently than the AI used by products such as ChatGPT and Claude. Google and Microsoft have also invested in the startup, and a deal with Nvidia could be announced as early as next week. The acquisition would expand Nvidia’s AI capabilities and follow its purchase of DeepMap in 2022.

Sources

NOTE:

This news brief was generated using AI technology (including, but not limited to, Google Gemini API, Llama, Grok, and Mistral) from aggregated news articles, with minimal to no human editing/review. It is provided for informational purposes only and may contain inaccuracies or biases. This is not financial, investment, or professional advice. If you have any questions or concerns, please verify all information with the linked original articles in the Sources section below.

AI infrastructure semiconductors data centers chip manufacturing memory chips high-bandwidth memory (HBM) AI demand cloud computing networking equipment semiconductor industry TSMC Nvidia Applied Materials Cisco Ciena CoreWeave Nebius AI investment AI startups AI regulation

Comments

Loading...