Search Market News — WSJ, Bloomberg, NYT, FT, CNBC, Reuters, Yahoo Finance Archive
Loading analysis...
Loading Truth Social...
Loading live updates...
Nvidia (NVDA) is joining rival AMD (AMD) along with partners Google (GOOG, GOOGL) and Meta (META) to establish a new open specification for data center design.
Called the Open Data Center Specification v0.7, the standard serves as a blueprint for building data centers, covering everything from the physical layout of the buildings to cooling and power architecture.
The chip giant also said it's teaming with Google, Meta, and Microsoft (MSFT) to create the OCP Grid Resilience Specification, which serves as a standard for how data centers respond to power disruptions, ensuring they can rapidly return to full operation.
According to Nvidia, the two specifications provide everyone from system makers to colocation providers and data center operators with a set of predefined templates for building data centers without having to contend with a slew of different customer designs.
Data centers are incredibly complex facilities that require large amounts of power and cooling to keep the chips inside from overheating and failing. And adding custom designs and quirks can slow construction timelines.
Open standards could change that, allowing builders to bring new data centers online faster. And the faster data centers go up, the more chips, systems, and networking components Nvidia can sell.
At the same time, data centers are facing increasing pushback from communities across the US, with some towns and counties imposing moratoriums or outright bans to stop companies from building them.
New York has enacted a statewide, one-year moratorium on data center construction, and Texas has halted new projects until it can determine the impact that connecting more power-hungry sites will have on its electrical grid.
President Trump has pushed back against any resistance to data centers, saying they'll be moneymakers for states.
AMD (AMD) is diving further into the agentic AI race with its new AMD Helion data center rack server.
Unveiled at the OCP Global Summit in San Jose, Calif., on Monday, Helion is designed to be outfitted with various configurations of CPUs and GPUs based on customers' needs.
The Helion is intended to serve as a reference design for AMD partners and clients, providing instructions for building the platforms.
The company says the server can be fully equipped with its 6th-generation EPYC CPUs for handling agentic AI tasks and servers, or configured with a collection of CPUs and its Instinct MI350P GPUs, which would serve as an all-purpose AI platform.
"Agentic AI is not one single workload. Agents are required for enterprise applications to running code and coordinating complex workflows," AMD senior vice president and general manager of compute and enterprise AI, Dan McNamara, said in a statement.
"To do that takes optimized compute and infrastructure that can adapt to different workloads. With AMD Helion, we are bringing AMD EPYC performance, open software and flexible rack designs together so our partners can build systems around customer needs and make efficient use of available power," he added.
The move allows AMD to better compete with rival Nvidia (NVDA), which offers up its own rack server reference designs, including its NVL72 servers.
AMD, like the rest of the chip industry, continues to benefit from the AI boom. The company topped analysts' estimates on the top and bottom lines during its most recent earnings report and provided a better-than-anticipated outlook for the current quarter.
Earnings per share rose 245%, while revenue climbed 50% to $11.5 billion.
Nvidia, however, continues to dominate in terms of sales, bringing in $96.2 billion in the prior quarter.
AMD stock, though, has outperformed Nvidia's over the last year, climbing roughly 182% year to date. Nvidia stock is up 25% in the same time period.
Microsoft (MSFT) unveiled its Azure Maia 300 AI chip at the OCP Global Summit 2026 in San Jose on Monday, seeking to reduce reliance on Nvidia (NVDA) and AMD (AMD). Microsoft says multi-gigawatt deployments beginning next year will make Maia its primary AI inference system for products and services. It claims Maia 300 delivers 40% more tokens per megawatt than leading GPU-based systems, while a 96-chip rack provides seven times Maia 200’s rack-scale token throughput. Other hyperscalers and AI labs also offer or are developing custom chips, but the article says demand for different workloads means hyperscalers will continue buying Nvidia processors.
The S&P 500 has risen 14% to record highs in 2026, but over 200 constituents have declined as gains concentrate in chipmakers and AI-related heavyweights. Reuters’ analysis of LSEG data shows the median stock gained barely 4%, while a quarter fell at least 11%; it excludes 18 companies removed this year. Sandisk and Micron more than tripled. Technology rose 32%, contributing over 10 percentage points to the benchmark’s gain, while energy climbed 42%. The Federal Reserve’s September rate hike and expectations of further increases hurt demand for industrials, consumer goods sellers and real estate companies. Fewer than half of current constituents trade above their 200-day moving averages.
Six-month-old, Sequoia-backed Nuvacore is raising hundreds of millions of dollars at roughly a $2.5 billion valuation to develop a data-center CPU, two people familiar with the fundraising said. The round remains open, and its size and valuation could change. The startup has no product yet and says it will develop core chip functionality before choosing an architecture, aiming to optimize data-center and AI computations. Founded by Gerard Williams, John Bruno and Ram Srinivasan, Nuvacore announced Sequoia-led seed funding earlier this year. Its fundraising comes as demand for CPUs grows alongside AI computing needs.
Anthropic co-founder Tom Brown is building business and political alliances ahead of an IPO that could value the company above $2 trillion. He resolved a two-and-a-half-week June shutdown of two models by assuring Trump administration officials that Anthropic had strengthened safeguards. Brown also negotiated a deal announced in early May to rent SpaceX computing capacity for $1.25 billion monthly, despite Elon Musk’s criticism of Anthropic. He maintains discussions with senior administration officials, but the Pentagon’s security-risk designation and ban on Anthropic tools remain in place, with conflicting court rulings on the ban.
The Nasdaq Composite slid on Thursday as new details about OpenAI spooked investors, putting the artificial intelligence trade under pressure.
The tech-heavy Nasdaq was down 1.25% to close at 27,193.34, while the S&P 500 declined 0.47% to 7,765.36. The Dow Jones Industrial Average added 51.77 points, or 0.1%, to end at 51,231.64.
Shares of key tech names came under pressure after OpenAI's annualized revenue was revealed to be short of the amount previously signaled. Shares of Oracle were down more than 5%, while chipmakers such as Nvidia and Advanced Micro Devices lost nearly 3% and almost 4%, respectively.
Adding to the downbeat sentiment, oil prices spiked also after President Donald Trump said he doesn't want to make a deal with Iran to end the war, while the U.S. is reportedly preparing for "massive bombing" in the Middle East. However, he later said that the U.S. won't attack Iran before the midterm elections early next month, moving oil off its highs. Brent crude was up 4.07% to $104.28 per barrel. West Texas Intermediate futures, meanwhile, advanced 3.64% to settle at $91.49.
Traders also watched the moves in rates after the 10-year Treasury yield and 30-year yield scaled to fresh 24-year highs this week. The benchmark 10-year yield was last down 5 basis points at 5.227%, while the 30-year yield was down 6 basis points at 5.601%. Those moves came after a solid 30-year auction by the Treasury.
"At the end of the day, I think it's inflation pressures, primarily driven by energy, that have shifted the Fed's disposition and have in a lot of ways shifted the long end of the curve as well," said Ross Mayfield, Baird investment strategist. "If we announced a deal with Iran tomorrow that was structurally sound and removed a lot of the tensions from the area, and oil dropped from $90 a barrel to $70 a barrel, yields would come down meaningfully."
Yields and oil have stoked volatility in equities of late, as concern grows that higher energy prices will keep inflation elevated and force the Federal Reserve to further raise rates.
While Mayfield said equities have been "pretty resilient" overall in the face of these pressures, with the S&P 500 and Nasdaq notching fresh all-time highs this week, he still doesn't think it would "take much of a move higher" on the long end of the yield curve to spell "more significant headwinds" for stocks. He believes yields could stabilize in the 5% to 6% range.
"We are in the process of the market adjusting to these new yields," he added. "In a lot of ways, higher yields are reflective of stronger nominal growth and slightly higher inflation, and if that's the case going forward, then again, it makes the argument that this kind of higher-for-longer or normal-for-longer rate environment is probably here to stay, barring a big recession where the Fed cuts rates to zero."
Palantir Technologies was one of the few stocks that closed higher on Thursday, with shares gaining 2% after the name received an upgrade at Goldman Sachs. The firm called for more upside ahead as the total addressable market could be "setting up for another step function change" as a result of the shift to sovereign AI and bespoke applications.
AI stocks fell Thursday after CNBC confirmed OpenAI reported roughly $50 billion in annualized revenue at September’s end, below the previously reported $68 billion. An unnamed source said the higher figure included partners’ gross revenue; OpenAI also reported third-quarter run rate growth of 77% overall and 107% for its enterprise business. Nvidia fell 3%, Oracle nearly 6% and CoreWeave nearly 8%. OpenAI, valued at $852 billion, confidentially filed its prospectus in June, with executives signaling a possible 2027 debut. CNBC previously reported preliminary funding discussions could raise around $30 billion, though the amount could change and no term sheet has been finalized.
Box (BOX) CEO Aaron Levie said on X that AI agents will require enormous computing capacity, rapidly growing inference volumes, and supporting infrastructure, describing the buildout as still early. He was responding to estimates of computing needs for Meta’s (META) Muse AI agent. Nvidia (NVDA) and AMD (AMD) reached record highs, while Marvell Technology (MRVL) raised its revenue outlook by billions of dollars. Creative Planning data showed Nvidia, Apple (AAPL), and Microsoft (MSFT) represent over 21% of the S&P 500, its highest-ever three-stock concentration. Yorkville Ives partner Dan Ives called AI the largest technology buildout in history.
The Nasdaq composite led major indexes lower Thursday afternoon. Here's what's moving markets today:
Rising oil prices: Oil prices are elevated, despite falling from their highs after President Trump said on social media that the U.S. won't attack Iran before the midterm elections.
Falling Treasury yields: Treasury yields fell after Trump's post and after a $22 billion auction of 30-year government bonds priced about in line with expectations.
Chip maker selloff: Shares of chip makers slid, with Intel, Marvell, Sandisk and Advanced Micro Devices all falling more than 4%. Nvidia was also lower, with the stocks taking a leg down after a Financial Times report raised questions about OpenAI's revenue outlook.
Wall Street investors favor chipmakers and AI infrastructure over betting on the leading AI agent, citing agents’ greater computing requirements and shifting product leadership. Meta’s Muse drew strong downloads and reviews, helping its shares gain 27% in September. The Philadelphia Stock Exchange Semiconductor Index has risen 84% this year. Bloomberg Intelligence expects AI infrastructure net income to grow 63% in 2027 and revenue 54%, up from July-end forecasts of below 48% and 32%, respectively. Citigroup analyst Atif Malik estimates the addressable CPU chip market will expand at a 60% compound annual rate through 2030.
Taiwan Semiconductor Manufacturing Co. posted a 51% quarterly revenue increase to NT$1.49 trillion ($46.7 billion) for the three months ended in September, according to Bloomberg’s calculations, exceeding analysts’ average estimate of NT$1.46 trillion. The growth reinforced the outlook for AI demand despite investor concerns about rising debt and regulatory setbacks. TSMC, a supplier to Nvidia and Apple, saw its shares reach a record this month. It is considering further expansion in Texas because of strong AI demand, after pledging $265 billion for an Arizona campus. North American customers contributed more than 75% of first-half sales.
Microsoft unveiled the Nvidia-powered Surface Laptop Ultra, starting at $2,599 and shipping soon, about $400 below an entry-level 16-inch MacBook Pro. Executives said Wednesday in San Francisco that it handles local AI prompts twice as fast as Apple’s MacBook Pro M5 and generates videos six times faster. Nvidia’s new Windows laptop chip supports general-purpose and AI computing; Dell and HP have also developed devices using it. Microsoft CEO Satya Nadella said users would eventually expect AI tools to combine local hardware and cloud services. Gartner said earlier this year that rising memory costs would likely delay AI PCs becoming a majority of shipments until 2028.
More gains may be in the cards for Taiwan Semiconductor (TSM) even as the stock has enjoyed a massive year of moving up and to the right on the charts.
"Another round of consensus earnings upgrades likely to come [on TSMC]," Citi analyst Atif Malik wrote in a note on Wednesday.
Malik lifted his earnings estimates for the company for the next two fiscal years. Two areas stood out in Malik's report:
He expects another strong year of results in 2027: "Supported by robust AI compute demand, potential upside from agentic AI, and the new networking cycle of CPO [co-packaged optics], we expect >40% year over year revenue growth for TSMC into 2027E," Malik said. "Along with full-year N2 contribution and more AI compute workloads moving to N3 from 2H26 and N2/A16 into 2028, we expect to see continuing consensus earnings upgrades as tight wafer foundry supply supports sustained demand growth. We believe TSMC's capex could increase over the coming years supported by strong revenue outlook."
The AI semiconductor market is likely to double next year: "We expect TSMC's AI revenue growth to nearly double next year considering most of its AI semiconductor customers are expecting accelerating growth," Malik explained. "TSMC's largest customer, Nvidia has guided to a doubling of its order intake and over 70% year over year revenue growth in FY2028E driven mainly by ongoing supply constraints. AMD has guided to over 80% CAGR in its Data Center AI business, while Broadcom expects its AI semiconductor revenue to almost double in FY2028E."
Taiwan Semiconductor shares hit a record high this week. They are up 61% year to date.
The company reports earnings on Oct. 15.
Stocks are sitting at record highs, but under the hood, it's the AI megacaps driving the gains. The rest of the market is flat or lagging.
"The breadth of the rally has narrowed," UBS's Ulrike Hoffmann-Burchardi warned on Tuesday.
The strategist noted that the percentage of US stocks trading above their recent price trends has fallen to levels last seen in March, a sign that fewer stocks are participating in the rally.
The trend has been gaining steam over the past month. At the end of last quarter, just 25% of S&P 500 constituents were currently trading above their 50-day moving average, according to Citadel Securities research.
While the S&P 500 (^GSPC) gained 2% in Q3, the S&P 500 Equal Weight fell 2%, and the Russell 2000 (^RUT) fell 7%.
The touchstones of the AI trade are keeping the indexes higher.
Nvidia (NVDA), Apple (AAPL), and Microsoft (MSFT) represent over 21% of the index, according to data from Creative Planning.
Nvidia touched an all-time high on Tuesday, along with peer AMD (AMD). Apple and Microsoft are both sitting near their record highs, as seen on Yahoo Finance's AlphaSpace chart.
In recent sessions, Taiwan Semiconductor (TSM), optical AI player Lumentum (LITE), and cybersecurity stocks have hit record or recent highs as demand for AI infrastructure and agentic AI surges.
UBS warned against overexposing to a narrow group of leaders and advised broadening portfolios.
"We retain strong conviction in the AI growth story," Hoffmann-Burchardi wrote. "But the increasing concentration of market gains reinforces the importance of managing risk through a broadly diversified equity portfolio."