How well do you understand your competition?

Not Another DeepSeek Article

Six months ago I wrote about how Meta’s open source Llama software changed the game for B2B companies. Every technology cycle has three acts, the means, the models and the applications. Meta makes its money from applications and wants to speed as many customers as possible through to that stage.

The strategy worked, with Nvidia’s Jenson Huang saying he wants to do for robotics what Meta has done for enterprise AI. Many of Nvidia’s own models are built on Llama and designed to make it easier for companies to build profitable applications. But there is a catch.

Llama is expensive to run making it an enterprise technology, rather than one for small companies. Meta understands that big businesses want their own models, and operating costs of half the level of GPT4o were considered attractive. Not any longer.

This is not another article about DeepSeek. What matters is that model running costs are collapsing, which is necessary for AI to be widely adopted. Reasoning models do a lot of their heavy lifting during run time and to be relevant to small businesses, this must be cheap.

The Journey and the Destination

At the height of the boom, railroad investment in Britain was 7% of national output. Once the track was laid, the competition shifted to building efficient engines. But the real money was made in the real estate and businesses built in newly accessible locations. Trains gave Britain’s industrial revolution a second wind.

The parallel journey with AI is from chips, to foundational models and then to applications. The big winners are not yet known and success will be in AI-enabled services rather than standalone AI products. It’s similar to going green with proven products, rather than building sustainably then looking for a market.

If you are in the services business, which of the three stages do you serve? OTAS used APIs to deliver quantitative analysis to traders. The big money was in the basis points clients shaved from their costs. Like OpenAI, OTAS was vulnerable to reverse engineering of its technology and price competition.

OTAS was bought by Liquidnet. Its core business was running an exchange and the technology to trade blocks in the dark. The big money is still made by the clients.

The services business model is always the equivalent of the rail or the engine, delivering significant value to clients. That is why they buy from you. There will be copycats and prices fall as technology matures. Vertical integration of the means – the exchange – and the models – trading technology – protects Liquidnet’s business even as competition rises.

Competition does not always meet you head on. Liquidnet reported the share of trading volume, executed in blocks of 10,000 shares or more of S&P 500 companies, fell 77% in the five years to 2022. The algorithms of trading firms became the core competition, rather than rival exchanges. Continuous trading promised lower risk adjusted execution costs over a day. Block trading also reduced as the absolute value of share prices rose. A $1 million trade involves far fewer shares after prices have risen a long way.

Ultimately, traders want to trade shares. They must finish their orders and demonstrate they did so in a low risk, cost effective way. The asset management industry involves executing trades on and off exchanges, to provide investment products to their clients. The big money, outside of hedge funds, is made by clients.

Appropriate Pricing

When you run a business you must understand where you sit in the industry. It’s likely you are at the third stage for one company and the first for another. When you map that out, you will understand where the competition is and get a better insight into the appropriate pricing of purchases and sales.

Questions to Ask and Answer

1. Does your business make the rails for any industry?

2. What are the rails that your business depends on to operate?

3. What is a left-field event that would reduce demand for your services?

When you are ready there are three ways I can help:

1. Schedule a call to chat or ask questions about AI with my tech team.

2. Resolving Team Conflicts: A free email course tackling an issue that no one teaches you as a manager.

3. The Profit Through Process Planner: My flagship course on how to design and invigorate a business that scales. I share 30 years of experience of researching, investing in and running companies, intermingled with the science and stories of business.