Who’s making money with GenAI?
Accenture booked over $600M this quarter ($2.4B annualized) in Generative AI
To put this number in perspective, in the full year of 2023, OpenAI made $1.6B in revenue.
Most enterprises see promising early results of GenAI experiments and plan to increase their spend anywhere from 2x to 5x in 2024 to support deploying more workloads to production.
This is a massive opportunity for founders building AI startups who:
1) Build for enterprises’ AI-centric strategy while anticipating their pain points
2) Build Productized services which capture this new wave of investment
A Revenue Multiple measures the valuation of an asset, such as a company, relative to the amount of revenue it generates
Control and Customizability are why enterprises care about open-source and self-hosted LLMs.
Hugging Face has consolidated its place in this ecosystem.
Llama became a de facto standard for LLMs, as Meta was the first to make a strong LLM source available and is reaping its benefits.
Competitors are following suit, with Google, Apple, Mistral, Stability, Databricks and others open-sourcing their models.
By January 2024, chatbots accounted for 46% of LLM apps.
- Customer Support Chatbots and
- Recommendation systems (with chat interface refining recommendations based on user input)
These are two of the primary customer-facing use cases that enterprises are looking at.
The biggest impact has been productivity gains - “A Dollar Saved is A Dollar Earned”
Above are the prevailing use cases, built in one of three ways
- Internal teams building solutions
- Consultancies leveraging productized services
- Product companies selling one-fit solutions
ROI is mainly due to the increased productivity generated by AI.
This will change, AI agents are about to explode.
2024 is the year where we go beyond simple wrappers and start building agentic workflows to accomplish complex tasks using AI.
Most AI apps today are workflow apps, which execute a sequence of actions to get to the final state.
SaaS apps are workflow wrappers on top of a database to put it simply.
we needed tons of workflows so we built tons of SaaS apps.
they all took user inputs and executed workflows — workflows that were the means to the end that customers wanted.
Now, with AI agents that can execute workflows (and soon will be able to execute fairly complex workflows flawlessly)
When AI agents will deliver the desired end results, there will no need for SaaS apps to design workflows.
so, agents come in, workflows go away.
Then Who has the moat?
There needs to be three core competencies in business for a strong moat
- Product innovation
- Operational Excellence
- Customer Intimacy.
For Generative AI apps, it can be Data or Infra.
If you are an LLM Infra company like Langchain, Pinecone or LlamaIndex, you have a good chance.
In 2023, enterprises spent $1.1B+ on the LLM infra stack — making it the largest new market in generative AI & a massive opportunity for startups
The total funding raised is $10B+ as of March 2024 (excluding $1B+ raises)
Or a company with data that no one else has or is hard to replicate, like Character AI.
You can also do great with GPT wrapper products. FormulaBot, SiteGPT are good early examples.
In the mobile app segment, apps that provide companionship like Character AI are doing phenomenal
Overall, ChatGPT takes the win here, being the all-rounder tool it is.
In conclusion,
- If you are just starting up and have limited funds in hand, you should bet on AI services then Productized AI Services and Consultancy
- If you have funds and experience, validate your idea and build a strong product on a specific use case with the end customer always in mind
- If money and talent aren’t your problems, then invest gargantuan amounts in R&D, get hold of very high-quality data, hoard compute, explore exotic and new architectures both in software and hardware
Sources:
