|
|
|
Dec 04, 2025
|
|
|
|
|
|
Happy Thursday! Salesforce's earnings provide an insight into its AI business. Meta poaches top Apple designer Alan Dye. OpenAI buys startup Neptune. Nvidia says its server can boost the performance of top open-source AI models by a factor of 10.
|
|
|
|
Salesforce reported 9% higher revenue in its fiscal third quarter ended Oct. 31, a marginal slowdown on its second quarter result, but the company raised its full year growth projection marginally. Its results also show that its flagship AI product, Agentforce, is growing rapidly but at a slightly slower rate than a quarter earlier. The enterprise software giant said Agentforce hit an annual recurring revenue rate of $540 million in the quarter, up 330% on a year earlier. In the second quarter, Agentforce’s ARR was $440 million, up 400%. Together with Salesforce’s Data 360 data product, Salesforce’s AI-related revenue hit an annual revenue rate of $1.4 billion. Salesforce has battled questions all year about whether businesses were ready to buy AI software, and whether its business was threatened by
competition from new AI services. Its AI-related revenue remains small for Salesforce overall. Its quarterly revenue was $10.3 billion and the company projected full year revenue of around $41.5 billion, up between 9% and 10%. That growth rate includes a contribution from the recent acquisition of Informatica. Salesforce has gradually raised its projection for fiscal 2026 revenue growth from between 7% and 8% issued earlier this year.
|
|
|
|
Meta is hiring one of Apple’s top designers, Alan Dye, who ran the company’s user interface design team, Bloomberg reported. Longtime designer Stephen Lemay will take his spot at Apple, the news outlet reported. Apple’s design team has suffered from regular bouts of turnover since the company’s legendary designer Jony Ive left in 2019. Many of those former staffers followed Ive out the door to his design firm, LoveFrom. Apple’s most recent iPhone operating system design, dubbed Liquid Glass, which makes menus and other parts of the user interface appear translucent, was
heavily criticized for poor legibility. There is also broader churn in Apple’s executive ranks. Earlier this week, the company announced that its top AI executive, John Giannandrea, would be stepping down. And last month, the company’s chief operating officer, Jeff Williams, who also oversaw the design team, retired.
|
|
|
|
OpenAI has agreed to buy Neptune, a startup that makes tools for monitoring and analyzing progress during AI model training. Terms of this transaction weren’t disclosed but OpenAI is paying less than $400 million in stock, according to people with knowledge of the deal. OpenAI’s Chief Scientist Jakub Pachocki said it plans to integrate Poland-based Neptune’s tools into its research workflow to understand better how models learn. The AI lab is using its highly-valued stock to fund acquisitions. The company, which was recently valued at $500 billion, acquired a small startup called Software Applications Inc. in October, product development startup Statsig for $1.1 billion in September and io, an AI hardware startup co-founded by former Apple designer Jony Ive,
for more than $6 billion in May.
|
|
|
|
Nvidia said one of its most powerful AI servers, which enables 72 chips to work in unison, can boost the performance of some of the world’s top open-source artificial intelligence models by a factor of 10. The announcement comes as Nvidia faces intense competition from other AI chips including Google’s tensor processing unit and Amazon’s Tranium chips. In a blog post Wednesday, Nvidia said its GB200 NVL72 system can dramatically improve the performance of leading open-source AI models such as Kimi K2 Thinking, developed by Chinese startup
Moonshot AI, and DeepSeek’s R1. Those top open-source models use a technique known as mixture of experts, which generate responses using far less computing power by routing queries only to the most relevant part of the overall model. Developing large-size models that use this technique while delivering high performance has been a challenge, and the NVL72 system solved that problem by tightly integrating its chips with its networking hardware and custom software, Nvidia said.
|
|
|
|
Database firm Snowflake said Wednesday its revenue rose 29% to $1.21 billion in the quarter that ended Oct. 31, beating its growth forecast for the quarter by 3 percentage points. CEO Sridar Ramaswamy attributed the growth to increasing usage of Snowflake’s database products by companies running AI applications. Snowflake also slightly raised its growth projections for product revenue in the fiscal year that ends at the end of January to 27% from 25%. Despite the upbeat results, Snowflake’s stock fell 8% in after-hours trading as the company projected a slightly lower growth rate for product revenue in the current quarter compared to the last one. Snowflake also said it will spend $200 million over several years to pay for AI models from Anthropic to power AI features in its database software, though the company
didn’t lower its projected gross profit margin for the current fiscal year. Snowflake recently started selling AI it says can automate customers’ workplace tasks ranging from resolving IT tickets and producing dashboards to handling customer service, joining a crowded field of software providers offering similar AI products. Ramaswamy said on Wednesday that Snowflake has surpassed $100 million in annualized revenue from its AI products, fulfilling a goal that he set internally earlier this year.
|
|
|
|
Morgan Stanley has held preliminary talks with investors about offloading some of the credit risk associated with its data center lending activities, Bloomberg reported. The bank is considering selling what’s known as a significant risk transfer product tied to loans it’s made to companies building physical infrastructure for artificial intelligence, the outlet reported, citing people with knowledge of the matter. SRTs can take a variety of forms but generally transfer a portion of losses from loan defaults to third-party investors, freeing up banks to do more lending Morgan Stanley’s potential SRT deal is an example of how banks are hedging the risks associated with the funding boom for AI data centers. Lenders are growing more cautious about a flood of debt tied to Oracle data centers, in particular, The
Information reported. A Morgan Stanley spokesperson declined to comment on the Bloomberg report.
|
|
|
|
| |
