Bullpen.News From Edwin Dorsey
Get Daily Recap Feedback

Wednesday, October 28, 2026

Back to today

From X

@SouthernValue95 1 click

SouthernValue on X

Ramp data shows VC-backed companies are spending 10x more per employee/month on AI vs. PE-owned and other companies. A bearish interpretation could be: token spend is being driven by VC-backed companies, making demand vulnerable to swings in capital cycle. Valid! However, on a longer timeframe, I think this is quite interesting as a leading indicator for the direction of software companies (VC-backed companies are mostly software or service-as-software companies). New companies are being built with a larger ratio of tokens vs. humans. Look at Anthropic itself, which pays 1/6 as much on intelligence as its customers (since it sells tokens marked up at 85% GM): Anthropic is spending ~9x more on compute than human salaries despite having access to tokens at cost. Cash payroll is annualizing ~$2B for a company with ~$60B of run-rate revenue in 1H26. Anthropic employs just 5k people. Wild efficiency in human capital. I suspect that the great software companies of the next decade, both new and incumbent, will insert AI into every aspect of their products so that using them feels like magic, and work happens seamlessly. Building a presentation in Claude feels like magic: I tell it what I want to put on a slide, and it appears like magic. As a user of Ramp, their products feel the same way: I make a purchase, take a picture of a receipt, and the expense is logged into my accounting system appropriately without the need for a human to type into a blank box. I think we will see most software become AI-native or be replaced by software that is. This is a significant opportunity for incumbents and startups alike. And I think this is a huge source of future token spend that will go to $AMZN and $GOOG where most software companies are built (on AWS and GCP) and where AMZN and GOOG are building the platforms for software companies to build AI capabilities into their products with choice around compute and model, embedded safety and governance, and unlimited scale. Amazon CEO @ajassy said on a podcast this week that AI spend is currently a "barbell-shape" with insatiable compute appetite of the AI labs on one side, and corporates using AI for cost savings on the other. In the middle is tokens embedded into "core workloads" which he says is growing but small today, but he thinks in 5 years might be the biggest category of token spend.
@convequity 1 click

Convequity on X

We think AI power users are overestimating how far enterprise adoption has progressed. Anthropic’s S-1 says over half the Fortune 500 were $100k+ annual customers as of September. Among those publicly traded customers, Anthropic spending averaged under 0.1% of total expenses in Q3 2026. That measures spending on one provider, so it cannot establish whether their AI deployments are succeeding. But the expectation gap is familiar. When you spend all day using the best models, it is easy to assume the rest of the economy is close behind. Building and running an enterprise’s own AI applications is a much harder undertaking. In Part 2 of our recent $NVDA report, we argued that successful adoption remains concentrated in a few exceptionally capable organisations, with Palantir ($PLTR) and coding agents the clearest examples. For Nvidia, adoption through clouds and model providers still leaves compute demand concentrated in intermediaries with the scale and incentive to develop their own chips. It needs more enterprises buying its compute directly. This is why we continue to see Palantir as a strategic enabler. The ability to help enterprises put AI into production remains scarce, even as access to capable models spreads. From Convequity’s “Nvidia at 16x Earnings: Bargain or Value Trap?” (Pt.2), 20 Oct 2026.

Investor Letters

All letters

Danaher’s robotics era

Danaher’s new CEO paired a Q3 beat and raised EPS outlook with a 2027 4%-5% organic-growth starting point, while pitching AI-enabled…

DHR $163B
Best Anchor Stocks 1 click

ORCL Update

Oracle rose about 12% intraday on October 28, 2026 as AI infrastructure and cloud-demand sentiment improved and credit-spread concerns…

ORCL $438B
Cestrian Capital Research 2 clicks

Meta Earnings, Muse and the Frontier, Musebook

Meta’s earnings hinge on whether investors back Zuckerberg’s Muse vision and the compute spending required to build it.

META $1.6T
Stratechery by Ben Thompson – On the business, strategy, and impact of technology. 1 click

Anthropic Deep Dive

Anthropic’s $80.7 billion first-half loss was mostly fair-value marks, while $16.3 billion revenue and compute economics showed a $974…

AMZN $2.6T
CV's Substack 7 clicks

My Last GFL Update - Hopefully

GFL’s take-private structure hinges on BC Partners rolling 60% of its stake to preserve debt portability under the 40% permitted-holder…

GFL $16B
310 Value’s Newsletter 5 clicks

Publications we follow

Every publication on this site, hand-picked.

129 publications

Newsletters 109

YouTube 4

Blogs 15

Seeking Alpha 67 of 266

Nominate a publication

X accounts we follow

Every account on this site, hand-picked.

91 accounts
Nominate an account

Investor letters we follow

190 firms

Sectors

Ordered by how often our editor reaches for them.