Pinecone's vector database just got a lot cheaper

Today: how Pinecone hopes to carve out space for the standalone vector database amid an industry stampede into the sector, Satya Nadella does Davos, and the latest funding rounds in enterprise tech.

Pinecone's vector database just got a lot cheaper
Photo by Jason Mitrione / Unsplash

Welcome to Runtime! Today: how Pinecone hopes to carve out space for the standalone vector database amid an industry stampede into the sector, Satya Nadella does Davos, and the latest funding rounds in enterprise tech.

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

Planting the seeds

As new enterprise technologies emerge, buyers looking for tools to manage those technologies are usually faced with two options. There are always a handful of startups that have grown up around that emerging tech and promise the kind of focused attention that established vendors racing to duplicate the technology can't, but they don't have the deep relationships and dependable track record with enterprise buyers that those established vendors already enjoy.

Pinecone is one of those startups, perhaps the buzziest company to emerge from the rush to adopt vector databases amid the generative AI boom. On Tuesday it announced that its flagship database is now "serverless," a word that takes on an expanded definition every year but is generally understood to mean "managed infrastructure."

  • The new version of Pinecone means "developers don’t need to provision, manage, or even think about infrastructure," and it also introduces usage-based billing, the company said in a blog post.
  • "In order to build remarkable GenAI applications, developers need an even easier and cost-effective solution for searching through massive, ever-growing amounts of vector data," said Edo Liberty, co-founder and CEO of Pinecone, in that post.
  • Pinecone serverless comes with new integrations into several popular platform services, such as Pulumi and Vercel, as well as widely used AI development tools, including Anyscale and Cohere.
  • “We’ve been working on it very hard for a year and a half now — this has been our most ambitious project," Liberty told VentureBeat.

Vector data is very important for large language models, which have to sift through a mountain of data to work their magic.

  • Vectors give data a directional quality, which makes it much easier for the model to understand the relationships between different words, phrases, and sentences.
  • Combined with a technique called retrieval augmented generation (RAG), vector databases help reduce the hallucination problem famously associated with LLMs.
  • However, like everything else associated with the generative AI craze, this is a very expensive process.
  • Pinecone said its new serverless data could reduce its operating cost by up to 50x because the new architecture separates key tasks like reading existing data, writing new data, and storing it all into separate activities, which "means you don’t have to pay for compute resources when you’re not using them," according to the company.

That's a massive reduction in operating costs. Should that experience pan out across Pinecone's user base, the new architecture — which it touted as an "industry first" — could help Pinecone compete against a crush of giants coming into this category.

Pinecone raised $100 million for its technology last year, a year when it became much harder to raise funds for enterprise tech startups compared to the recent past.

  • Venture-backed startups drove an outsized amount of innovation in enterprise tech over the last two decades, thanks in large part to incumbents that stuck their heads in the sand and pretended nothing had changed.
  • Today's incumbents have learned from those mistakes, and Pinecone's traction in enterprise tech will be an interesting story to follow as generative AI moves from something enterprises talk about doing to something they actually do.

It's Davost wonderful time of the year

While in Davos, Switzerland, for this year's World Economic Forum, on Tuesday Microsoft CEO Satya Nadella announced a sweeping partnership with U.K. mobile giant Vodafone and discussed his company's headlong plunge into generative AI and Sam Altman's orbit.

As part of their deal, Vodafone agreed to spend $1.5 billion building AI services on top of Microsoft Azure over the next ten years, allowing it to shut down several data centers. In return, Microsoft will prop up the telco's Internet of Things division when it is spun out later this year.

In a wide-ranging interview with Bloomberg's Brad Stone, Nadella also insisted that "stability" at OpenAI is more important to Microsoft than having a controlling stake in the startup company that it relies upon to drive a significant amount of its product roadmap. “I feel very good about the construct we have,” he said. “I feel at the same time very capable of controlling our own destiny.”

Enterprise funding

Quantinuum landed $300 million in new funding from JPMorgan Chase, valuing Honeywell's quantum computing company at $5 billion.

Infield launched with $3 million in seed funding for its open-source dependency management software, which could help companies fix issues with their software supply chains.

The Runtime roundup

Meanwhile at Davos, OpenAI confirmed that it is working with the U.S. military on some cybersecurity projects.

Snyopsys bought Ansys, which makes software for designing and testing products, for $35 billion to complement its chip design software.

Synx bought Helios, a small startup working on observability tech, to add that fast-growing technology to its cloud security software tools.

Cybersecurity Dive published a great deep dive (sorry) into the MOVEIt disaster, one of the worst ransomware incidents ever reported.

Thanks for reading — see you Thursday!

Great! You’ve successfully signed up.

Welcome back! You've successfully signed in.

You've successfully subscribed to Runtime.

Success! Check your email for magic link to sign-in.

Success! Your billing info has been updated.

Your billing was not updated.