Three Economic Reasons GenAI Innovation Will Only Grow
Is it time to unleash your killer model?
As we noted in our recent Harvard Business Review article, Where to Get Started with Generative AI, we are seeing significant adoption of GenAI especially in work that creates, manipulates and improves Words, Images, Numbers and Sounds (what we call WINS Work). In addition to breakthrough performance of the many GenAI models, ease of using them and largely free initial access, there are massive economic reasons that Generative AI is only beginning. (If you want a complete view of the LLM Market, see our brand new research: The Corporate Buyer’s Guide to LLMs).
The Top 6 Vendors Have Enormous Cash Flow Enabling Massive Ongoing Investment
Despite the recent drama about OpenAI, it’s vital to keep perspective on the absolute size of this wave of GenAI technology. When I hear people worrying, about whether Microsoft can afford these large language models I have to laugh a bit as Microsoft, by itself, produces almost $2B in Earnings Before Interest Taxes Depreciation and Amortization (EBITDA) per week! That’s just under $400 million every working day. To put this in perspective one government expert friend of ours recently told us that the Russian government collects about $370 billion in taxes in a typical year. Together the top six firms generated $428 Billion in their 2023 fiscal year, most of them ending in June of 2023. (See the graph below.)
All of them are investing in GenAI models either within their own firms or in massive commitments to innovative firm. Amazon has committed $4B to Anthropic, and built out there Bedrock AI hosting and development environment. Microsoft has invested internally – where they are building about three dozen co-pilots for their products as well as committing billions to the newly reconstituted not-for-profit/for profit OpenAI combo-organization. Nvidia is not only investing in massive chip development (while minting cash) but also their own language models and software tools. Meta is spending billions to build their open-source models which by 2024 should be on Qualcomm chips and embedded in smartphones, enabling at the edge GenAI.
Not only do they have the money, but early indications are that they are reaping incremental rewards from their existing customer base. Microsoft’s marketing test of a $30/month/user price point for its co-pilot offering in the office suite seems to be working. Google is trying too. If Microsoft can get its price point to stick – and it provides GenAI based co-pilots for all its major applications, it will likely raise revenue significantly and give them even more stickiness inside the enterprise. Google worries that if they are not competitive their 90% market share of search may be under threat. The land grab is on among some of the most cash rich organizations ever created.
This portends well for all of us as this battle royal will create great new technologies and the fact that each firm starts with different strategic assets, we may benefit from one firm giving away the razor to sell the blade and vice versa. For example, Meta will push open source and edge while Microsoft is likely to push cloud and integration in their existing enterprise stack as they chase Amazon. Moreover, the fact that there are six means that it is less likely from a game theory standpoint that you will have competitive signaling and tacit price collusion (which is legal, as long as it’s tacit) as I believe we have in internet provision by cable companies.
The Top 6 Firms Have Staggering Valuations That Gives Them Enormous Market Power
The combined market capitalization of Apple, Microsoft, Google, Amazon, Nvidia and Meta are market capitalization as of October 2023 was $10.1 trillion, compared to the GDP of Germany and Japan which combined is $8.6 trillion. We know that market capitalization and GDP measure completely different things, but comparing these gargantuan valuations to nation states is.
The cynic might say, these market values are a bubble and due to irrational exuberance and a hope for the “greater fool theory” in which a greater fool comes along and buys your speculative stock from you in the hope it will go higher – as many did in the dot com rally. But these massive valuations are nothing like the internet bubble where Drugs.com and many others had revenue, high investments, and generated no cash. Here, the giants of the industry are fighting it out with rivers of money -- as noted above, $428 billions of EBITDA in 2023 alone. If we look simply at their 2023 together, which is not cash, but it’s close to cash, we see another amazing set of numbers. In 2023, with most fiscal years ending at the end of June, they generated $428 billion in EBITDA.
Significant Market Power Likely Facilitates a Friendly Regulatory and Risk Environment.
Republican John Sherman of Ohio, younger brother of Union Civil War General, William Tecumseh Sherman and author of the famous anti-trust act that bears his name said, “If we will not endure a king as a political power, we should not endure a king over the production, transportation, and sale of any of the necessaries of life.” The big 6 have vast lobbying and government relations efforts as well as first class attorneys. We do not see any 21st century Teddy Roosevelt in the wings coming in to break up these formidable giants. If we look at how Google enhanced their street view product going across many territories capturing data that was illegal to capture, they had lawsuits in over 22 countries that they successfully delayed, contested and where necessary settled. Microsoft and other vendors are indemnifying those who use their generative ai environments from any IP risks.
Given the vast market power, cash generating ability and political influence – and lack of a new trust buster in the American political landscape, there’s no question in my mind that the presence of vast Generative AI capability is already begun and going to grow. If there are increasing political battles it will be important to separate possible histrionics from candidates from implemented policy.
Unleashing the Killer Model
Given this vast hive of innovation, I think a central strategic question to ask is: Is there a killer model to be created in my industry? Consider what proprietary data assets might your firm have that can be turned into a model that has economic scale and improved quality. For example, Paige.ai has partnered with Microsoft using its 4,000,000+ digitized microscopy slides and over a billion images to create what they hope will be the world’s best digital pathology/oncology model. Likewise, Mayo Clinic has announced a collaboration with Google to use Google’s Gen App Builder to create Generative AI models that are HIPPA compliant. There is no guarantee that these efforts will be successful, but if they are they may radically lower the marginal cost to serve and have vast distribution potential. Many types of WINS work may go the way of classified advertising. Large scale providers with very low marginal cost.
Every firm should be asking themselves -- are there data or transactions or interactions that we have stored or have superior or even exclusive access to? Would that data be even more valuable if it were combined and analyzed by Generative AI the way Paige.ai expects their project to deliver?
The best place to start is to ask what are the biggest sources of data and transactions we create or access? Some firms like Ensemble Health Partners has vast amounts of medical and medical billing data. Can they make an industry scale model? Firms like Kaiser Permanente who have some of the most comprehensive patient, treatment and insurance payment data in the world may be able to create some of the most useful medical models.
If you firm has such advantaged access to transactions, data, customer service, asset locations, etc., etc, now is the time to begin thinking about if you can build a proprietary model to create strategic advantage. You may decide to fund this effort in house. Alternatively, you could consider partnering with one of the six giants as Paige.ai and Mayo have, to leverage the technology provider’s cash, market power and expertise.
In short, GenAI will continue to be with us and grow at a rapid rate. WINS work tasks need to be reengineered to be competitive, and every firm must look at its overall data assets to see if there is a way to grow a model that could create a new growth path and competitive strategic advantage.