Microsoft announced earnings yesterday & the data painted a brilliant picture for the future of AI.

Greater than 30% annual growth in back-to-back quarters is sensational for a $100b run rate business. Microsoft is projecting similar for next quarter. The AI subset is on a $13b run rate, more than double last year.

Azure other cloud services revenue grew 31%. Azure growth included 13 points from AI services, which grew 157% year-over-year, and was ahead of expectations even as demand continued to be higher than our available capacity… In Azure, we expect Q3 revenue growth to be between 31% and 32% in constant currency driven by strong demand for our portfolio of services.

Darwin’s finches sprinted to evolve when food sources changed. For decades, nothing—then sudden transformation. AI follows the same pattern & it’s accelerating.

Every morning, I wake up wondering what breakthrough will propel the ecosystem forward.

Last week, it was DeepSeek v2. This morning, a Hugging Face researcher announced that he could induce reasoning on the major advance of the last three months in one of the smallest models, a 3B parameter model.

A microwave that writes its own recipes. A smart watch that crafts personalized workout plans. A ticket kiosk that negotiates refunds in natural language. This isn’t science fiction - it’s 2025, & DeepSeek just made it far more affordable.

The Chinese AI company released two breakthroughs: V3, which slashes training costs by 90+%, & R1, which delivers top-tier performance at 1/40th the cost. But the real innovation? They proved that sometimes simpler is better.

There are two opposing forces in the world of data: an overall consolidation within the modern data stack & a massive expansion driven by AI capabilities. AI is rewriting every rule about what’s possible with data in 2025.

Here are Theory’s Top Themes in Data in 2025 with the full presentation at the bottom.

1. The Great Consolidation. After a decade of expanding complexity in the modern data stack, companies are looking to dramatically simplify their architectures to drive better results. Buyers we speak to say, “Do not sell me another tool.”

   As a result, we are seeing consolidation on individual cloud data warehousing platforms Snowflake & Databricks, where most enterprises have picked their dominant architecture. There is also a wave of consolidation within BI favoring collaborative BI tools that balance centralized & decentralized control like Omni. There is a race to command more & more compute within these consolidated platforms because the majority of revenue & ultimately profits resides there.

   The office of the CFO continues to apply pressure to drive more ROI on core data & AI, which is a surprise given 2010-2022 data budgets grew unabated & the fervent interest in AI. In particular, this has pressured the cloud data warehouses, & customers are looking for novel architectures to dramatically reduce cloud data warehouse spend where possible. 50% cost savings are possible with newer transformation architectures like SQLMesh / Tobiko Data.

2. Scale-Up Architectures. New cloud data storage formats are increasing in importance although slower than expected in terms of adoption because they lack the relevant enterprise tools. However, in the long term, this creates a rise in workload-specific query engines like MotherDuck & Datafusion.

   These query engines are typically scale up rather than scale out. This means developers can start on their local machines & take advantage of the phenomenal computing power of their MacBooks to handle all the vast majority of workloads.

3. Agentic Data. If the IT department in the future is the HR department for AI agents, we should expect that data will be transformed just as much as every other team. Historically, there has been a divide between software engineering & AI/ML teams, & that will change. Many of the software design principles, like virtual environments, will come to data.

   We should expect the vast majority of SQL queries to be executed by AI. To make these queries accurate, data modeling becomes an absolutely essential technology to eliminate hallucinations & guarantee quality. In addition, data observability tools like Monte Carlo will become increasingly important as data not only feeds the core BI & analytics layers but also the AI systems that are key parts of every production application, both internal & external.

   That AI should drive some pretty significant efficiencies that parallel the 25% to 50% productivity gains across Google, Microsoft, & ServiceNow, or the $275 million in cost savings of Amazon migrating from one version of Java to another. This should free budgets for new initiatives.

   Smaller models will dominate within the enterprise with the sweet spot somewhere between 10 billion to 70 billion parameters, because of a 600X difference in inference cost & very similar levels of accuracy. If the wave of innovation from DeepSeek is any indication, there’s significantly better performance & a continued deflation & inference cost over the short & intermediate term.

Slide 1

Clearing: While data world consolidates, capabilities have exploded with AI.

Content:

• AI is rewriting every rule about what’s possible with data
• Those two forces in tension will make for an exciting 2025

Slide 2

Clearing: My name is Tomasz Tunguz, founder and general partner at Theory.

Content:

• I’ve been investing in data for the last 17 years and have worked with companies like Looker, Monte Carlo, Hex, Omni, Tobiko Data and Mother Duck
• I founded Theory, a venture firm managing $700M with the idea that all modern software companies will be underpinned by data and AI
• We run a research-oriented firm, formed by 200 buyers of data and AI software

Transition:

Over the weekend, a small Chinese hedge fund turned star AI research outfit launched DeepSeek R1, a new massive open-weights model with state-of-the-art performance, trained on a shoestring budget.

Just how much interest is there in this advance?

I analyzed R1 downloads on Ollama, and I recorded my steps to perform this analysis with AI using speech, an AI model, & a developer environment. See the video below if you’re curious how I did it.

A few days before the deadline, I’d find myself in that familiar pool of anxiety : staring at a blank digital canvas, clock ticking, knowing the next 20 hours would dissolve into a blur of bullet points, chart creation, & late-night pixel-alignment.

Time to make a presentation.

With AI, I’ve slashed the time it takes to build a presentation from a full day down to just a few hours

I start with the essence - three key ideas or stories that will resonate with my audience.

The most successful AI startups of 2024 shared an unlikely secret: they didn't rely on proprietary datasets. Instead, they leveraged synthetic data to outmaneuver competitors who were still chasing exclusive data partnerships and expensive labeling operations.

The numbers tell a compelling story. Synthesis AI grew 410.6% last year while Datagen raised $72M – the largest funding round in the synthetic data space. Meanwhile, companies burning millions on human data labeling watched their unit economics deteriorate as synthetic alternatives delivered 500-1000x cost reductions.

On Thursday January 30th :at 5:30 Pacific time, Office Hours will host Maggie Hott, GTM Leader at OpenAI.

This Office Hours session will be held in person at SHACK15 in San Francisco & is for early-stage founders.

Maggie will dive into mistakes founders make building GTM for AI companies — ranging from hiring practices to pipeline building & pricing strategies—and provide actionable insights on how to avoid them.

Despite holding a staggering $370b war chest in 2025, tech giants aren’t racing to acquire companies – they’re too busy building their AI empires, one data center at a time.

Microsoft hasn’t acquired a company since January 2023. Two of Google’s reported acquisition overtures, Wiz & Hubspot, were scuttled.

This focus elsewhere creates an opportunity for mid-market players to lead the M&A wave of 2025.

As of January 2nd, these major acquirers hold $370b in cash & short-term equivalents. These levels represent near-historic highs over the past 15 years.¹