Kirkland ibuprofen is the same molecule as Advil. Same dosage, same FDA requirements, same therapeutic effect. It costs 80% less.

AI has its generic drug moment. DeepSeek V3 matches GPT-5.2 on most benchmarks. It costs 90% less. OpenAI & Anthropic generated $22 billion in 2025. Chinese AI labs generated $1.8 billion. The ratio : 12:1. Pricing explains the gap. Chinese AI API prices collapsed 90% in 2024. US frontier models average $3.38 per million input tokens. Chinese models average $0.48.

Leveraged software companies running on leveraged infrastructure. When AI compresses software revenue, the stress doesn’t stop at equity. It cascades into debt.

BDC assets hit $475 billion in Q1 2025. Software comprises 23% of Ares Capital, the largest BDC.

Shares of Blue Owl, Ares, & KKR dropped 9%+ on Tuesday. UBS estimates 35% of BDC portfolios face AI disruption.

BDCs (Business Development Companies) are publicly traded private credit funds. They became the primary lenders to software over the last decade as private equity sponsors bought software companies with debt. The sponsors’ thesis was simple. Software revenue is durable, so lenders will accept 4-6x EBITDA leverage. AI is already writing code, conducting legal research, & managing workflows cheaper than legacy SaaS. The recurring revenue backing those loans is the target. Anthropic’s autonomous legal agents announcement sent LegalZoom & Thomson Reuters down 12%, echoing ChatGPT’s impact on Chegg & Stack Overflow. AI can vaporize software revenue.

If we needed another exclamation point on the tremendous growth opportunity, NVidia’s earnings punctuated the euphoria with gusto. Revenue grew 88% in a quarter, nearly doubling. The company had projected $11b & exceeded projections by $2.5b or 26%.

Earlier today, New Relic announced its sale to Francisco Partners & TPG for $6.5b.

The acquisition is notable for two reasons.

First, it accelerates the momentum within the technology buyout space. At its current pace, technology buyout volumes of venture-backed technology companies will tie or exceed the ten year high, charted in 2022 of about $20b.

The $1m ARR figure used to be the benchmark in 2018 & early 2019 for raising a Series A. But the data shows how much the market differs from a few years ago.

Series A round size standard deviation has grown by between 4-5x in 4 years. A Series A used to mean a single flavor. Today, like a Neapolitan ice cream, Series As can mean a $1m round, a $23m round or a $110m round.

I think of marketing teams as hedge funds. Marketing teams develop a portfolio of different strategies to acquire leads.

Some days, content marketing works. A post challenging like this one () hits HackerNews or a journalist covers the company.motherduck.com/blog/big-data-… Other days when content strategies struggle, a witty paid ad campaign entices some clicks & form fills. On some rainy Tuesday, the weekly webinar attracts an unusually promising audience.

Like an investment portfolio, any individual strategy may thrive one day & suffer the next.

Yesterday, the Enterprise Tech 30 List was revealed. Congratulations to all the winners (especially MotherDuck, Hex, & Omni)!

I analyzed the headcount patterns within these companies to shed light on three questions : 1. How are these top companies changing their headcount through the downturn?
2. What percent of headcount is in product & engineering?
3. What percent of headcount is in sales & marketing?

A product manager today faces a key architectural question with AI : to use a small language model or a large language model?

This is my current mental model of when to choose a large or small model : When to choose a large model :

- time to ship is critical : many of these models are available via API, requiring formatted data as an index or vector database - which an engineer can achieve within a few hours for a working beta.

Great companies transform a technology innovation into a go-to-market advantage.

In the last five major cycles (internet, social, mobile, cloud, web3), startups seized the new technologies of the era to create a distribution advantage.

What will it be for AI? There are a few emerging possibilities :

1. OpenAI plug-in architecture enables companies to build products integrated into a new search & discovery experience. With millions of users to address, a novel plug-in experience could catapult a startup.

I’ve been tinkering with LLMs.
First, I created a chatbot using all my blog posts. Then I created a model to produce blog entries based on my writing.
As I went through this process, I asked myself some questions along the way : Does the ML model ingest my posts & keep them? What if I’d like to remove them from the logs or training set or output for others?

Which database generated a -$1.8b loss & within a year produced $130m in profits? Ethereum.

I asked ChatGPT about the numbers 1 & 4. Which one is bigger? Sometimes, 1 was bigger. Other times, 4 was bigger.

Founder-led sales are sales engineering sales, not account executive sales.

As a company scales, founders transition sales to others. The default behavior is for founders to pass sales to account executives. What if they passed it to sales engineering? When a startup is born, founders lead sales. Design partners, pilots, or founder-led sales - they have many names - these customers work with a startup to solve a problem.

The Azure ML business is roughly on a $900m run rate (by my rough estimation), with customers growing 10x. Azure’s AI products could achieve a multi-billion dollar run rate by the end of the year, a data point which underscores the idea that AI grow GDP 1000x more than the PC. I’m watching public company earnings to identify early trends in the software market to inform startups’ plans for 2023. Yesterday, Microsoft & Google announced earnings. Amazon, Cloudflare, & Mongo announce later this week.

As machine learning becomes core to every product, engineering teams will restructure.

In the past, the core engineering team & the data science/machine learning teams worked separately. The core engineering team ships the product & focuses on reliability. For most companies, the data science team analyzed data & wrote machine learning models to support the business functions : sales, marketing, customer support. These teams operated downstream of the data warehouse.

Over the past week, I put ChatGPT to the test, enlisting it as copy-editor for a forthcoming blog post. So, does a robot copy-editor work in practice? Here was my process:

1. I dictated a blog post outline into a Google document
2. I prompted ChatGPT to revise the post for clarity & pasted the text
3. I copied ChatGPT’s response, edited it some more & consulted ChatGPT for three rounds just like John McPhee

In August, I wrote about why @ExploreOmni's technology reinvents BI once more. The Omni founding team hails from Looker & Stitch. They’ve set out to solve the problems customers faced with the previous generation of BI.

Today, I can show you what this means in practice. Anyone who has managed a larger BI deployment has faced the challenge of managing hundreds, perhaps thousands of metrics. As more users calculate figures, consistency across teams becomes a company-wide challenge.

I’m watching public company earnings to identify early weaknesses in the software market. Yesterday, Microsoft announced earnings.

Microsoft is a Mirror of What We Can Expect for SaaS in 2023

The earnings transcript highlights the 5 major trends in software of 2023. 1. Enterprises Have Slowed their Spending, Decelerating Further in December:

Growth is down 15 percentage points or about a third in a year.

The current wave of layoffs, a difficult component of the innovation boom/bust cycle, differs from the previous years’ dynamics.

B2B companies have reduced headcount to a greater extent than at any time since 2020. In the last three years, B2C startups’ ratio of layoffs have dwarfed B2B layoffs. In 2020, B2C companies cut 8.8x the number of B2B employees. 3.8x in 2021, & 6.9x in 2022.

Year-to-date in 2023, the figure is 1.6x, just 60% more.

“The Startup M&A Market Fell 94% Year over Year - But One Segment Thrives”, a post from last week, elicited the same question from many readers.

Just who are the most active acquirers for smaller takeouts? To answer the question. I reviewed the roughly 560 US software acquisitions under $400m since January 1, 2020.

Just how big is the current web3 B2B SaaS total addressable market (TAM)? In the last six months, 103 web3 companies generated revenue on-chain, the smallest of which recorded a few hundred dollars of sales & the largest, Ethereum, tallied $401m.

44% of these companies produced less than $0.5m. But 41 companies produced between $5-25m.