Cerebras Beats NVIDIA Blackwell in Llama 4 Maverick Inference
SUNNYVALE, Calif., May 28, 2025--(BUSINESS WIRE)--Last week, Nvidia announced that 8 Blackwell GPUs in a DGX B200 could demonstrate 1,000 tokens per second (TPS) per user on Meta's Llama 4 Maverick. Today, the same independent benchmark firm Artificial Analysis measured Cerebras at more than 2,500 TPS/user, more than doubling the performance of Nvidia's flagship solution.
"Cerebras has beaten the Llama 4 Maverick inference speed record set by NVIDIA last week," said Micah Hill-Smith, Co-Founder and CEO of Artificial Analysis. "Artificial Analysis has benchmarked Cerebras' Llama 4 Maverick endpoint at 2,522 tokens per second, compared to NVIDIA Blackwell's 1,038 tokens per second for the same model. We've tested dozens of vendors, and Cerebras is the only inference solution that outperforms Blackwell for Meta's flagship model."
With today's results, Cerebras has set a world record for LLM inference speed on the 400B parameter Llama 4 Maverick model, the largest and most powerful in the Llama 4 family. Artificial Analysis tested multiple other vendors, and the results were as follows: SambaNova 794 t/s, Amazon 290 t/s, Groq 549 t/s, Google 125 t/s, and Microsoft Azure 54 t/s.
Andrew Feldman, CEO of Cerebras Systems, said, "The most important AI applications being deployed in enterprise today—agents, code generation, and complex reasoning—are bottlenecked by inference latency. These use cases often involve multi-step chains of thought or large-scale retrieval and planning, with generation speeds as low as 100 tokens per second on GPUs, causing wait times of minutes and making production deployment impractical. Cerebras has led the charge in redefining inference performance across models like Llama, DeepSeek, and Qwen, regularly delivering over 2,500 TPS/user."
With its world record performance, Cerebras is the optimal solution for Llama 4 in any deployment scenario. Not only is Cerebras Inference the first and only API to break the 2,500 TPS/user milestone on this model, but unlike the Nvidia Blackwell used in the Artificial Analysis benchmark, the Cerebras hardware and API are available now. Nvidia used custom software optimizations that are not available to most users. Interestingly, none of the Nvidia's inference providers offer a service at Nvidia's published performance. This suggests that in order to achieve 1000 TPS/user, Nvidia was forced to reduce throughput by going to batch size 1 or 2, leaving the GPUs at less than 1% utilization. Cerebras, on the other hand, achieved this record-breaking performance without any special kernel optimizations, and it will be available to everyone through Meta's API service coming soon.
For cutting-edge AI applications such as reasoning, voice, and agentic workflows, speed is paramount. These AI applications gain intelligence by processing more tokens during the inference process. This can also make them slow and force customers to wait. And when customers are forced to wait, they leave and go to competitors who provide answers faster—a finding Google showed with search more than a decade ago.
With record-breaking performance, Cerebras hardware and resulting API service is the best choice for developers and enterprise AI users around the world.
For more information, please visit https://www.cerebras.ai/.
View source version on businesswire.com: https://www.businesswire.com/news/home/20250528123694/en/
Contacts
pr@zmcommunications.com
Hashtags
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
Yahoo
16 minutes ago
- Yahoo
Is Super Micro Computer Stock Headed for $15 or $70 After Giant Post-Earnings Plunge?
After putting accounting controversies and Nasdaq delisting fears stemming from the delayed filings in the rearview mirror, Super Micro Computer (SMCI) had been on a strong recovery run this year. SMCI stock once again regained traction as investors bet big on its leadership in the artificial intelligence (AI) hardware space, growing demand for next-gen servers, and close partnership with chip giant Nvidia (NVDA) — all of which painted a promising growth story. But the narrative took a sharp turn in the wrong direction after the company dropped its fourth-quarter earnings report on Aug. 5. Shares of SMCI sank more than 18% in the following trading session, as Super Micro missed Wall Street's expectations on both revenue and earnings. Even guidance fell short, sparking fresh concerns about slowing growth and rising competition. More News from Barchart Robinhood Stock Seemingly Can't Be Stopped in 2025. Is It Too Late to Buy HOOD Here? Dear Ford Stock Fans, Mark Your Calendar for August 11 Cathie Wood Is Buying Shares of This Little-Known Ethereum Treasury Company. Should You? Get exclusive insights with the FREE Barchart Brief newsletter. Subscribe now for quick, incisive midday market analysis you won't find anywhere else. Investor sentiment around SMCI stock has clearly taken a hit, and analysts' opinions are now deeply divided. On one end, the most bullish target on Wall Street suggests that SMCI stock could soar to $70, while on the other side, the most bearish sees it dropping all the way to $15 per share. So, with expectations stretched so far apart, can SMCI defy the doubts and reclaim its momentum? Or is there more downside still to come? About Super Micro Stock Super Micro Computer specializes in building high-powered, energy-efficient servers to support surging demand from AI, cloud services, and data centers. With operations across the U.S., Taiwan, and the Netherlands, the company has built a solid international footprint and currently holds a market capitalization of $27.8 billion. After sliding last year under the weight of negative headlines, SMCI stock came roaring back in 2025, hitting a fresh high of $66.44 in February. Even after pulling back some 32% from that peak, the stock remains up an impressive 50% so far this year, well ahead of the broader S&P 500 Index's ($SPX) modest 8.5% gain year-to-date (YTD). However, the comeback rally appears to be facing turbulence lately. SMCI is once again under pressure following its latest earnings miss, with shares tumbling 19% in just the past five trading days. A Look Inside Super Micro's Q4 Earnings Super Micro's fiscal 2025 fourth-quarter earnings report released earlier this month came as a letdown to investors, missing Wall Street's expectations on both the top and bottom lines. Revenue for the quarter rose 7.4% year-over-year (YOY) to $5.8 billion but still fell short of the $5.9 billion analysts had expected. Adjusted EPS came in at $0.41, marking a notable 24% decline from the prior year and falling below the consensus estimate of $0.44, partly because of the impact from President Donald Trump's tariffs on goods imported into the United States. While the company pointed to tariffs as a key reason for the earnings shortfall, the miss stung even more given the backdrop of soaring demand for AI infrastructure, which had set investors' expectations sky high. Instead of a blowout quarter, the results revealed mounting cost pressures and thinner margins, casting doubt on Super Micro's ability to deliver strong profits even as sales continue to rise. A key red flag was the drop in gross margin, which slipped to 9.5% from 10.2% a year ago. The decline suggests the company may be facing pricing pressure or higher costs, likely tied to competitive dynamics or more expensive AI components. As of June 30, Super Micro reported $5.2 billion in cash and equivalents, alongside $4.8 billion in total bank debt and convertible notes. Unfortunately, the disappointment didn't end at Super Micro's Q4 financials. The company's guidance for Q1 of fiscal 2026 also fell short of expectations. Management expects revenue to land between $6 billion and $7 billion, while adjusted EPS is expected to be between $0.40 and $0.52. Both of these figures miss Wall Street's forecasts for $6.6 billion in revenue and $0.59 in EPS. Over the longer term, analysts tracking Super Micro project the company's bottom line to rise 19.2% YOY to $2.05 per share in fiscal 2026, then leap 21% annually to $2.49 in fiscal 2027. What Do Analysts Think About Super Micro Stock? After the disappointing Q4 showing, several Wall Street analysts struck a more cautious tone. For instance, JPMorgan analyst Samik Chatterjee lowered the firm's price target to $45 from $46, maintaining a 'Neutral' rating. Chatterjee pointed to capital constraints and customer hesitation as key reasons the company fell short of expectations. On the other hand, Bank of America Securities analyst Ruplu Bhattacharya slightly raised the price target to $37 from $35 while sticking with an 'Underperform' rating. The analyst noted that gross margins were pressured once again, this time due to inventory reserves for older products. Bhattacharya explained that many customers are opting to wait for Nvidia's upcoming B300/GB300 GPUs, which has weighed on current demand. Overall, Wall Street sentiment on Super Micro is tilting cautious, with SMCI stock earning a consensus rating of 'Hold.' Among the 16 analysts covering the stock, opinions are clearly mixed. Only four recommend a 'Strong Buy,' while three lean toward a 'Moderate Buy.' The majority of analysts take a wait-and-see approach, with six suggesting 'Hold.' On the bearish side, one analyst has a 'Moderate Sell' rating for SMCI and two assign a 'Strong Sell' rating, reflecting growing uncertainty around the company's near-term outlook. SMCI stock has a Street-high price target of $70, indicating nearly 54% potential upside. The lowest target of $15 suggests a steep 67% drop from current levels. Key Takeaways Super Micro's wide price target range, spanning from $15 to a Street-high of $70, underscores the uncertainty surrounding its near-term trajectory. While some analysts remain bullish on its long-term AI-driven potential, others are wary of execution risks and margin pressures. With sentiment this divided, the stock's path forward may remain volatile as investors wait for more apparent signs of sustained growth. On the date of publication, Anushka Mukherji did not have (either directly or indirectly) positions in any of the securities mentioned in this article. All information and data in this article is solely for informational purposes. This article was originally published on
Yahoo
2 hours ago
- Yahoo
Microsoft Stock Heads for the Clouds After Q4 Earnings. Is It Too Late to Buy MSFT Here?
Microsoft (MSFT), one of the "Magnificent 7,' recently delivered a strong quarter that gave markets plenty to think about. On July 30, shares jumped 9% in extended trading after the company posted fourth-quarter results that beat expectations on both revenue and earnings. Fiscal Q4 revenue guidance had already been set high by CFO Amy Hood, with a range between $74.7 billion and $75.8 billion. The actual results outpaced that range, marking Microsoft's fastest growth rate in over three years. Investors did not take long to react. MSFT stock surged past $550 after hours and continued to climb the next day. On July 31, it hit a session high above $555, pushing its market capitalization to around $4.1 trillion. That made Microsoft the second company after Nvidia (NVDA) to cross the $4 trillion mark. More News from Barchart Robinhood Stock Seemingly Can't Be Stopped in 2025. Is It Too Late to Buy HOOD Here? Dear Ford Stock Fans, Mark Your Calendar for August 11 Cathie Wood Is Buying Shares of This Little-Known Ethereum Treasury Company. Should You? Stop Missing Market Moves: Get the FREE Barchart Brief – your midday dose of stock movers, trending sectors, and actionable trade ideas, delivered right to your inbox. Sign Up Now! While the growth momentum is strong, Microsoft is not without its near-term challenges. The company continues to face data center infrastructure constraints as it scales up to meet the rising demand for artificial intelligence (AI). CFO Amy Hood had initially anticipated improved supply conditions by June, but that timeline has now shifted to December. Still, the pace of investment remains aggressive. At current run rates, annualized spending could exceed $120 billion, a 36% increase year-over-year (YOY). This level of expansion reflects Microsoft's long-term conviction in AI. With MSFT stock riding high on strong growth, let's see if there's still time to buy shares. About Microsoft Stock Based in Redmond, Washington, Microsoft is one of the most dominant names in global tech. With a current market cap of $3.9 trillion, the company's reach is everywhere. Microsoft's Windows operating system alone controls more than 70% of the PC software market. But that is only the beginning. Microsoft's product ecosystem spans cloud computing (Azure), productivity software (Office 365), developer tools, business applications, gaming, and more. From enterprise clients to individual users, the company has built a presence across nearly every category. Over the past 52 weeks, MSFT stock has delivered a 30% gain, comfortably beating the Nasdaq 100 Index's ($IUXX) 28% climb in the same period. Valuation remains high. MSFT trades at 34.2 times adjusted forward earnings and 13.8 times sales, placing it well above industry averages. Such a premium suggests investors are pricing in continued growth and AI-driven upside. Microsoft also remains a consistent dividend payer. It has increased dividends for 20 years in a row. Right now, it pays an annual dividend of $3.28, yielding 0.62%. The next quarterly payout of $0.83 is scheduled for Sept. 11 to shareholders of record as of Aug. 21. Microsoft Surpasses Q4 Earnings Microsoft's Q4 results posted strong revenue, strong profit, and strong momentum. Total revenue came in at $76.4 billion, up 18% YOY. The figure beat the $73.81 billion Street estimate, showing that Microsoft's core businesses are still expanding at a pace. Its Intelligent Cloud segment, which includes Azure, posted $29.9 billion in revenue, a 26% increase from the year-ago quarter. Productivity and Business Processes, home to Office and LinkedIn, brought in $33.1 billion, up 16% YOY. The More Personal Computing division — which includes Windows, Surface devices, Xbox, and search advertising — delivered $13.5 billion, a 9% YOY jump. Profit margins held strong, too. Gross margin hit $52.4 billion, rising 16.4%. Net income grew 23.6% to $27.2 billion. EPS grew 23.7% from the year-ago value to $3.65, beating the expected $3.37 per share. Cash and cash equivalents climbed to $30.2 billion, a 65% increase from last year. Microsoft's AI integration is showing up in the numbers. Copilot, the company's AI assistant for Microsoft 365 and Windows, now has 100 million monthly active users. That adoption is also lifting revenue per user across Microsoft 365 commercial cloud bundles. Looking ahead, analysts expect Microsoft's Q1 2026 EPS to rise 10.3% YOY to $3.64. Full-year EPS is forecast at $15.32, up 12.3%. Meanwhile, for fiscal 2027, the bottom line is projected to grow another 16.8% to $17.89. What Do Analysts Expect for Microsoft Stock? Analysts remain largely bullish on MSFT stock following the strong Q4 results. Daniel Ives from Wedbush raised his price target from $600 to $625 and kept an 'Outperform' rating. Ives highlighted Microsoft's ongoing strength in cloud and AI as the key drivers behind the revised outlook. Truist analyst Joel Fishbein also upped his target, moving from $600 to $650. He maintained a 'Buy' rating and pointed to accelerating Azure growth, AI adoption across services, and the broader shift to cloud-based infrastructure as reasons for the higher valuation. Analyst sentiment for MSFT remains optimistic with an overall rating of 'Strong Buy.' Out of 46 covering analysts, 39 maintain a 'Strong Buy,' five issue a 'Moderate Buy,' and two recommend to 'Hold' on the stock. MSFT stock's average price target of $617.42 represents potential upside of 18% from current levels. On the more bullish end, the Street-high target of $675 implies that the stock could climb 29% from current levels. Analysts still see more upside ahead, showing it's not too late to buy the stock. On the date of publication, Aanchal Sugandh did not have (either directly or indirectly) positions in any of the securities mentioned in this article. All information and data in this article is solely for informational purposes. This article was originally published on Sign in to access your portfolio
Digital Trends
2 hours ago
- Digital Trends
GT 210, RX 580, and more: Amazon's GPU bestsellers list is a time capsule
What's happened? Gamers are always chasing the best graphics cards, right? Not so much. The latest Amazon sales data showed that some truly ancient GPUs are still bought hundreds of times per month. New sales data from Amazon U.S., published by TechEpiphany, revealed some interesting insights into the current state of the GPU market. While the top-selling card in July was Nvidia's RTX 5070, the list contains many older GPUs from the last decade. The GT 710, a card from 2014, saw 1,050 units sold, putting it in the top 15 by unit sales. Similarly, the GT 1030, GT 730, and even the GT 610 each sold hundreds of units, despite being low-end or obsolete. High-end GPUs like the RTX 5090 and RTX 5080 dominate revenue but have far fewer total unit sales. The revenue from selling 1,450 RTX 5090s amounted to a whopping $3,863,674.50, which is still more than selling 6,050 units of the RTX 5070. The GT 210, a GPU that wasn't highly regarded even when it launched back in 2009, sold the same number of units as the recent Intel Arc B580. While the data is interesting, there's no way to verify whether every part of it is valid. 🔥 GPU Retail Sales Amazon US July '25 Total Units Sold: 41,800 AMD: 7,750 units (18.54% market share) Intel: 1,100 units (2.63% market share) NVIDIA: 32,950 units (78.83% market share) Total Revenue: $20,996,254.50 AMD: $4,284,973.50 (20.41% revenue share) Intel: $223,115.00… — TechEpiphany (@TechEpiphanyYT) August 6, 2025 This is important because: GPU prices are all over the place right now, and this sales data highlights that many PC owners may be priced out of modern GPU upgrades. Aging or ultra-budget cards like the GT 710 and GT 610 are still selling in volume, despite offering very poor gaming performance by today's standards. The trend suggests that entry-level PC builders or casual users are settling for cards that are underpowered or nearly obsolete. While many of the lower-end cards may have been bought for reasons other than gaming, they, too, have newer alternatives — possibly too pricey. Recommended Videos Why should I care? The data shows that flagship cards aren't the only way to go, and many people settle for less, given the current economic climate. If you're building or upgrading a PC, this data shows you're not alone in shopping on a budget. Finding a truly affordable GPU is still tough, and you might need to choose between old hardware and overspending. We do not advise buying these low-end cards for gaming — save up for something from the last two or three generations. Okay, what's next? If you need a new GPU, keep an eye out for GPU deals instead of buying something ancient. GPU prices have been high this whole year, but we're seeing signs of improvement. Unless you really need a discrete GPU, it's better to wait than to buy a $35 GT 210. AMD APUs with built-in graphics offer another alternative.
