In October 2024, OpenAI raised $6.6 billion, valuing the corporate at $157 billion. Quick ahead six months, and OpenAI is in talks of elevating $40 billion, at a valuation of $340 billion. This week, Anthropic introduced a $3.5 billion spherical, valuing the corporate at $61.5 billion. In December 2024, Elon Musks Grok raised $6 billion collection C, valuing the corporate at $75 billion lower than a 12 months from inception. Ilya Sutskever’s SSI is presently elevating cash at $20 billion valuation (pre-launch) and the record goes on.
On account of their quick development, each Anthropic and OpenAI grew to become two of essentially the most useful personal corporations in a brief time period. However is it sustainable?
The launch of Deepseek’s R1 mannequin on the twentieth of January 2025, was dubbed by many because the ‘Sputnik second’ within the AI race as a result of it challenged the undisputed US dominance within the foundational mannequin house up till that time. Deepseek, a Chinese language firm with out (seemingly) entry to the most recent Nvidia chips, claimed have spent solely $6M to coach and effective tune a reasoning mannequin that reached related efficiency benchmarks as OpenAI’s o1 mannequin. The prices later turned to possible be increased in actuality, however the potential of disrupting the dominance of US foundational fashions and the infrastructure they depend on (Nvidia) rattled the market and triggered an enormous tech dump and Nvidia’s lack of $650 billion in market cap in a single day. Since then different Chinese language corporations together with Alibaba and Bytedance have launched plenty of fashions difficult the US corporations, most are open-source and free or less expensive than their US equivalents.
Till now, nearly all of the worth in generative AI has been captured by the infrastructure layer: foundational fashions (LLMs), the cloud suppliers and the {hardware} infrastructure/ knowledge centres. Nevertheless now, the AI panorama is present process a major shift, prompting an important query: The place will the long run worth of AI accrue?
The thought is that going ahead the worth is about to shift from the infrastructure to the appliance layer and AI brokers. There’s precedent to this perception: throughout the first dot com bubble, browsers went public with enormous fanfare (see Netscape), however in a short time grew to become commoditised. It seems the worth of the Web was not within the instruments used to entry internet pages, however somewhat within the companies used (and paid) by companies and shoppers. Will that even be the case in AI?
The Wave of AI Commoditisation
The underlying developments suggests a transfer in direction of the commoditisation of basis fashions. Which means that the core AI fashions themselves have gotten more and more accessible and reasonably priced, levelling the enjoying subject for companies of all sizes. I’m unsure how shortly this can occur, if in any respect, however take into account the next arguments supporting this development:
- Reducing Prices: Coaching AI fashions is turning into cheaper and extra environment friendly. DeepSeek’s R1 mannequin, for instance, reportedly matches the efficiency of OpenAI’s flagship programs at a fraction of the price.
- Elevated Accessibility/ the rise of open supply fashions: Many foundational fashions are actually open supply or obtainable via cost-effective APIs, making AI integration simpler for startups and smaller enterprises. Working example: the CEO of Determine, the humanoid robotics firm publ
- Shift in Focus: The main target is shifting from the capabilities of the underlying fashions to how successfully they’re built-in into broader programs.
Each problem will be a chance
The commoditisation of LLMs might imply alternative for startups: going ahead, the bar for constructing AI-native merchandise is way decrease, enabling smaller corporations to combine AI and fostering innovation. As the worth shifts from the fashions themselves to how they’re applied, the aggressive benefit now lies in constructing dependable, autonomous AI brokers and embedding AI fashions into clever programs that may deal with real-world duties.
Corporations that construct domain-specific options or layer proprietary knowledge onto commoditised fashions can have an edge, spurring new functions throughout industries and finally resulting in market growth via value discount. This technological shift might require established gamers to adapt to keep away from being outpaced by extra agile rivals. The emphasis is shifting in direction of holistic programs that harness a mannequin’s outputs to automate duties and drive real-time decision-making.
One of many challenges of investing in AI is the wave of commoditisation. What appears novel on the time of investing determination will be absolutely commoditised a 12 months later (or sooner in some instances). That made it troublesome for some VCs to position bets – that’s not about to vary. Under are three examples of potential winners.
Vertical AI
Even with the commoditisation of foundational fashions, vertical AI is a big alternative as a result of it addresses complicated, industry-specific calls for with tailor-made functionalities, deep integrations, and specialised workflows.
Whereas foundational AI gamers deal with expertise, vertical AI applies these instruments to fulfill {industry} necessities. The aggressive edge lies in product and distribution, constructing defensible positions via {industry} alignment and efficient go-to-market methods. Vertical AI markets are likely to comply with a “winner-takes-most” dynamic, emphasising the significance of scaling quickly and capturing market share early. As a substitute of specializing in simply commoditised functions, vertical AI ought to deal with end-to-end workflows and combine tightly with present programs. Corporations can differentiate themselves by focusing on missed classes with complicated necessities and nuanced wants.
Examples from the Remagine Ventures portfolio: we invested in Trulux, develops AI-driven authentication and traceability options designed for the luxurious retail market in addition to Storywise, an AI platform remodeling the antiquated e book publishing {industry}.
Clever Programs and AI Brokers
AI brokers are well-positioned for achievement in a panorama of commoditizsing basis fashions as a result of they shift the main focus from the fashions themselves to the clever programs leveraging them. This evolution strikes from AI as a “co-pilot” to an “auto-pilot,” the place AI executes duties autonomously. AI brokers provide action-based worth by innovating and packaging duties into companies, significantly benefiting SMBs via entry to reasonably priced, refined companies through AI agent marketplaces.
These marketplaces, which flip software program into companies, are anticipated to dominate attributable to community results and the power to offer pre-vetted, on-demand AI companies. Success shall be additional pushed by constructing belief and explainability in AI brokers, enabling them to make clever choices, and creating AI-first organisations able to wide-ranging actions. I like to recommend studying Gigi Levy’s publish on the NFX weblog.
As well as, the infrastructure to construct these brokers is getting more and more extra strong in a fast timeframe. Madrona’s Jon Turow has a fantastic primer on this.
Examples from the Remagine Ventures portfolio: we invested in Aigency (stealth) which is creating agent to agent communication for the transactional internet in addition to Keewano, an AI co-pilot for cellular analytics, beginning with gaming and shopper apps.
AI Apps
AI apps are an excellent alternative as a result of AI-native merchandise are scaling sooner and interesting customers extra deeply. The introduction of extra superior fashions and capabilities have led to a surge in utilization. There’s growing fragmentation throughout completely different AI mannequin suppliers, permitting for specialization and providing creators extra tailor-made choices. Textual content-to-web app platforms are accessible to each technical and non-technical customers, altering who can create with AI, not simply who can use it. Many apps with decrease utilization have confirmed to be efficient at monetisation via cellular subscriptions.
A16Z simply revealed their third instalment of the highest 100 AI shopper apps – check out the panorama beneath.
In abstract
It’s too early to declare the commoditisation of basis fashions, and it’s secure to say, that despite the fact that these corporations could be overvalued, corporations like OpenAI, Anthropic, Grok and different LLMs have seen among the quickest waves of shopper adoption we’ve seen in latest instances.
The AI battle between the US and China will proceed to play a task (no less than throughout Trump’s presidency) and may very well result in a faster adoption of open supply fashions. What is definite is that for early stage traders, there’s an enormous alternative now, and that the following couple of years will proceed to be a rollercoaster journey when it comes to new improved fashions, and naturally, commoditisation.
I am a former normal associate at google ventures, head of Google for Entrepreneurs in Europe and founding head of Campus London, Google’s first bodily hub for startups.
I am additionally the founding father of Techbikers, a non-profit bringing collectively the startup ecosystem on biking challenges in assist of Room to Learn. Since inception in 2012 we have constructed 11 faculties and 50 libraries within the creating world.
