Intel and AMD team up to stop Arm

Welcome to Runtime! Today: Strange times make for strange bedfellows in chips, Google goes nuclear, and the latest funding rounds in enterprise tech.

(Was this email forwarded to you? Sign up here to get Runtime each week.)

A rivalry gets 86'd

It's hard to think of any technology first introduced in Studio 54's heyday that plays such a central role in today's enterprise tech world as the x86 instruction set, the marching orders for server chips built by Intel and AMD. Despite recent challenges to their hegemony, x86 chips still power the vast majority of cloud and on-premises servers in use today.

However, over all those years Intel and AMD tweaked x86 in subtle but incompatible ways to suit their own needs, which can cause problems for software developers. Putting aside decades of differences (including one major antitrust lawsuit) Intel and AMD announced the formation of the x86 Ecosystem Advisory Group Tuesday at the Open Compute Conference, promising to standardize the technology at the heart of their most important products.

With major partners such as Dell, Google Cloud, HPE, Lenovo, and Microsoft (and the noticeable absence of AWS), the group will work together "to shape the future of x86 and foster developer innovation through a more unified set of instructions and architectural interfaces," they said in a statement.
Intel invented x86 way back in 1978 with the development of the 8086 chip, and AMD was granted a license to become a second supplier to the growing PC market in 1984.
Right now, "x86 is led architecturally by Intel, and any alignment with AMD is achieved primarily between x86 ISVs," according to Patrick Moorhead, a longtime chip industry executive and analyst.
The group's goal is to ensure that everyone is working with one standardized version of x86 while also allowing outsiders to influence the future direction of the technology, rather than Intel more-or-less controlling the roadmap.

Several changes in the chip market led Intel and AMD to this historic partnership. One of them is the rise of alternative instruction sets such as Arm, which runs the chips found in nearly every mobile phone in the world.

Arm has made serious inroads into the data center over the last five to six years, starting with the introduction of AWS's Graviton processor in 2018, and now every major cloud provider has unveiled its own Arm CPU.
"Part of Arm's rise stems from rules in its contracts that all Arm chips be able to run all Arm software, regardless of who made the chip," Reuters noted.
Most software written for x86 chips can run on either Intel or AMD's designs, but each company has added custom instructions to their chips over the years that have "caused some inefficiencies and some drift in portions of the ISA [instruction set architecture] over time," AMD's Forrest Norrod told Tom's Hardware.

But while the need for a solution to those problems has been discussed for years, it took a substantial blow to Intel's dominance over the enterprise server market to get to this point. It's impossible to imagine the Intel of a decade ago, which commanded more than 90% of the market for data-center processors, agreeing to share directional control over one of its most important assets with its archrival and its customers.

One day before Intel introduced the new group, it began laying off 15,000 workers in response to the myriad financial problems it has rung up over the last several years, "the biggest downsizing in its five-decade history," according to The Oregonian.
Meanwhile, AMD has increased its share of the server market to 24%, "its highest server market share in decades."

Given that so much software running the world has been written for x86 chips, Intel will still be a major player in the data-center market for years to come if it can fix its financial situation and get its manufacturing and product-design strategies back on track.

And no matter how it got here, Intel deserves credit for doing the right thing by server buyers and software developers in trying to make sure x86 stays vibrant into its fifth decade.

Gone fission

Google Cloud joined the ranks of hyperscalers pledging a renewed commitment to nuclear power this week, unveiling a new power purchasing agreement with startup Kairos Power. Beginning in 2030, "this deal will enable up to 500 MW of new 24/7 carbon-free power to U.S. electricity grids and help more communities benefit from clean and affordable nuclear power," Google said in a blog post.

Kairos is working on so-called small modular reactors, or SMRs, which are a far cry from the massive facilities built in the 1970s and 80s. Last Energy, which is also working on SMRs, builds its reactors out of modules that can be chained together to add power gradually as needed.

Both AWS and Microsoft have signed nuclear power deals this year as the electrical requirements of the generative AI boom threaten to exhaust existing power supplies. However, "Google’s offer to take the power from Kairos’ SMRs offers genuine 'additionality,' as the phrase goes, by encouraging the construction of brand new reactors and, potentially, seeding a whole new industry in the U.S.," according to Bloomberg.

Enterprise funding

Decagon raised $65 million in Series B funding to build a larger team constructing its customer service AI agents, as the agentic AI frenzy heats up.

Galileo landed $45 million in Series B funding as it expands its platform for AI observability, hoping that AI developers need the same type of performance and reliability data that traditional developers crave.

Neuron7 scored $44 million in Series B funding to bring AI to customer service departments, which feels like the goal of two-thirds of enterprise software companies these days.

Xscape Photonics raised $44 million in Series A funding for its silicon photonics technology, which the company's backers believe will improve datacenter bandwidth while controlling costs and power consumption.

Port secured $35 million in Series B funding as it builds out its take on a modern DevOps platform.

Relyance landed $32 million in Series B funding for its data governance platform, which helps companies meet compliance requirements.

The Runtime roundup

Databricks agreed to use AWS's Trainium AI chips over a five-year period to power a new service that will help its customers develop their own AI models or refine existing ones.

The National Archives and Records Administration is trying to get employees on board with plans to use Google's Gemini AI technology for both public-facing and internal chatbots, according to 404 Media, and it's not going great.

Thanks for reading — see you Thursday!