Gossip (for nerds)
- Microsoft is having a rough month for PR. First, Gizmodo reports, “Microsoft Is Firing About 9,000 People Because Business Is Great.” As the company’s profits and stock hit historic highs, it let go of just under 4% of its workforce globally. Sales, customer service, and software engineering were all affected. Some developers are allegedly being replaced by the AI they helped build.
- For the newly jobless, a Microsoft executive suggests they console themselves with AI—he even gave them prompts.
- Microsoft isn’t the only cloud company cutting staff. AWS laid off hundreds in July. However, like Microsoft, it’s not just AI and automation driving these layoffs. While official reports are hard to find, there is much chatter here and elsewhere that the bigger influence is offshoring jobs to India and the Philippines, in part prompted by the shaky status of the H-1B visa program that brought those workers here to the US.
- Microsoft’s Maia AI chip is delayed by at least six months and even then, once it goes into production, it’s expected to fall short of NVIDIA’s Blackwell chip. I believe what you felt reading that is called schadenfreude.
- Analysts say Microsoft has so far captured the largest amount of generative AI spend in the market. Nearly 60% of CIOs plan to increase Azure spending next year, with 97% planning to adopt AI tools.
- What does the next fiscal year hold for Microsoft? The cloud giant will increase investments in migrating companies from VMware platform to Azure; boost partner funding and incentives across Copilot, Azure, and Microsoft 365; and will introduce several new partner specializations and designations.
Wheelin’ and dealin’
Meta and AWS are collaborating on a program that will provide six months of technical support from their engineers, plus our favorite form of currency, AWS cloud credits, to 30 startups building AI tools using Llama AI. I have no doubt Meta will use these tools responsibly. If you have a conversation about monkeys and then get a bunch of monkey videos on your Facebook page, that is purely a coincidence, I don’t know what to tell you.
Professional pivots
- The VP and general manager of generative AI at Amazon has peaced out to work for Siemens. He led AI product strategy at AWS but surely an AI agent can do his job, no? A fun fact in the article: In its efforts to lure OpenAI employees away, Meta has been offering $100M sign-on bonuses. That was not a typo.
- Snowflake snatched away the AWS Managing Director of Industries and Solutions. Also out the door after 17 years is Kevin Miller, the AWS Global Data Center VP.
New stuff
- New solutions and services from AWS this month include OracleDatabase@AWS, which is generally available in some US Regions; AWS Builder Center, a new online hub where AWS users can share ideas, access learning resources, join community programs, and vote on feature requests; and Kiro, which will help developers write code from AI.
- Ten years ago, I was getting these terrible ice pick headaches deep in my dome piece and I went to the doctor, and he was like it’s probably scalp pain. I said no, it isn’t. Then he said oh it’s probably anxiety. I said no, it isn’t. Then I got a CT scan and it was a brain tumor! True story! So when Microsoft says a study showed its AI can diagnose 400% better than doctors, not only do I believe them, but I would add to the list of other things that can out-diagnose many doctors: balls of lint, naked mole rats, rocks.
- While Microsoft AI is diagnosing you instead of a doctor, AWS can help with surgery. AWS, NVIDIA, and Johnson & Johnson have launched the Polyphonic AI Fund for Surgery, a grant program for building AI tools aimed at improving care before, during, and after operations. (Just don’t sprinkle any of that J&J powder on my organs. Or their hip implants. Or their birth control. Or Risperdal. Actually, yes to the Risperdal.)
- The AWS Summit in New York was home to several announcements, starting with Amazon Bedrock AgentCore. It’s a modular, enterprise-grade platform that enables organizations to deploy, manage, and scale secure AI agents at production-level. The underlying advantage is that it “bridges the critical gap between proof of concept and production for AI agents.”
- Other announcements of note: A $100M investment in agentic AI developments through the AWS Generative AI Innovation Center and a new AI agent Marketplace, in partnership with Anthropic. An update to Copilot Vision for Windows will allow the tool to see everything that’s on your screen, including browser windows.
- Microsoft, OpenAI, and Anthropic created The National Academy of AI Instruction to train 400,000 K-12 teachers on AI over the next five years, with a flagship campus in NYC.
- Project Rainier from AWS is a colossal, highly efficient “AI supercomputer” that links hundreds of thousands of custom Trainium2 chips across multiple data centers. This will give AI developers like Anthropic about 5x more power (isn’t that exactly what tech companies need right now?) to train next generation models.
- Blaxel thinks AWS and Microsoft don’t have the right infrastructure to handle AI agents at scale—and it’s already seeing significant traction. Palantir is also going toe-to-toe with AWS by offering an integrated platform for building sophisticated data and AI applications.
Ma’am, I’m going to have to call security
- SharePoint servers, including those belonging to governments around the world, healthcare providers, and energy companies, came under attack and a “broad level of compromise” this month when hackers took advantage of an “undisclosed digital weakness” discovered by a third-party security firm. At least 50 servers have been successfully compromised, leading Palo Alto Networks to describe this ongoing issue as a “high-severity, high-urgency threat.”
- To get around a government requirement that only US citizens with security clearances may access Defense Department data, Microsoft used a workaround: “digital escorts” that carried out tasks from engineers in China on sensitive Pentagon cloud systems—even though many of these escorts didn’t understand how to catch potential threats.