Helping data teams solve the world’s toughest problems
We are thrilled by today’s announcement from Databricks that it is acquiring MosaicML for $1.3 billion (see great Wall Street Journal coverage here). We have been with each company from their early days — we invested in Databricks in 2014, and co-led MosaicML’s seed in 2020 — and cannot imagine a better combination of forces to bring to the world AI that is high-performance, secure, affordable, and explainable.
Databricks is a data and AI company whose Lakehouse Platform is used by 10,000 organizations, including half of the Fortune 500, for the unification of data, analytics, and AI. For years, Databricks has addressed a complex problem: data warehouses are great for the centralization and analysis of structured data, but inefficient for unstructured data, while data lakes bring support for all data formats, but lack performance. Databricks’ Lakehouse combines the best elements of both, taking advantage of cheap, reliable cloud storage but making data more easily writable and readable. This helps enterprises make their data infrastructures more scalable, reliable, secure, and cost-effective.
MosaicML is built to put large-scale machine learning models within reach for more companies, and represents a compelling alternative to the use of gigantic, off-the-shelf large language models that cannot be controlled by the companies that use them. As I wrote last month, MosaicML has come up with incredible products just as LLMs are exploding in power and popularity: companies using Mosaic can deploy generative AI models quickly, easily, securely — and at one-fifteenth the cost of other comparable service.
MosaicML delivers the democratization of AI with performance and accountability, and enables every customer from the smallest startup to the largest corporations and governments with sensitive/mission-critical/“life-safety” applications to avoid the unaccountable, untransparent, unreliable serfhood model of AI that some giant tech companies want to force on their customers.
The acquisition enables the combination of Databricks’ data management with MosaicML’s machine-learning models, which in turn will allow even the world’s largest organizations to use their own data to build critical AI models at low cost — safely and accountably.
Both companies have been led by visionary leaders who inspire loyalty and extraordinary team performance and satisfaction with a unique combination of high EQ and high IQ. We offer our most enthusiastic congratulations to Databricks CEO Ali Ghodsi and his fellow cofounders, and to MosaicML CEO Naveen Rao and his cofounders.
DCVC is committed to shaping the development and use of AI — as its partners have for almost 30 years — and is on the lookout for more bold, thoughtful, principled teams like Databricks and MosaicML who are working to make computing accountable, flexible, and broadly accessible. Better AI can make a better world — and it does every day, across DCVC-backed teams delivering real world outcomes in the service of resilience, abundance, and the defense of democracies.
Matt Ocko is Co-Founder and Co-Managing Partner of DCVC.