If big techs are all buying the entire LLM market, is there any space left for startups?

In the 1980s and 1990s, the S&P 500 was dominated by a mix of industries, with leading companies like General Electric, ExxonMobil, and Walmart reflecting the economy. Since then, the scenario has shifted dramatically. Internet, software, and cloud played a huge role in lifting the “magnificent 7” — Apple, Microsoft, Amazon, Alphabet, Meta, Tesla, and NVIDIA — to the top of the index. Will AI do the same for the next generation of companies, and will we see new names at this top? TL:DR and my take: probably not.

Unlike GE, ExxonMobil, and Walmart, the Magnificent 7 seem to have learned lessons from its ancestors and started investing heavily in technology that could disrupt them. They didn’t think twice in the face of the innovator’s dilemma. Microsoft is championing a $13B partnership with OpenAI, Google integrated with DeepMind and invested $2B in Anthropic, and Amazon recently reinvested in Anthropic, totaling $4B.

Context on the LLMs race and big tech moves

Dominance in the LLM sector is pivotal because it underpins the broader GenAI ecosystem. The foundational role of LLMs means that applications spanning from content creation to business analytics, while potentially lucrative, will inherently depend on the capabilities of these underlying models. Essentially, the control over LLMs equates to influencing the power source of the entire GenAI-driven economy.

Significant investments and technological advancements from big tech companies predominantly shape the current LLMs market. With their deep pockets, extensive data repositories, and substantial computational infrastructure, these firms have established a strong foothold. Their dominance is further solidified by leveraging existing platforms for seamless distribution and integration of AI technologies, providing them an unmatched edge in efficiently scaling and deploying new AI-driven services and products. This combination of financial muscle, data access, and platform integration capabilities positions them as formidable leaders in the LLM space.

For instance, Microsoft’s colossal $13 billion partnership in OpenAI showcases the scale of commitment and the expected returns from the LLM market. Such investments are not only about developing new technologies but also about securing a competitive edge by incorporating LLMs into various applications, ranging from enterprise solutions to consumer services, such as Co-Pilot for Microsoft 365. The integration of LLMs into everyday business operations spans across industries, including healthcare and cybersecurity, indicating a broad utility spectrum that these models offer. This trend underscores a strategic shift towards utilizing AI to gain competitive advantages.

Big tech’s investment strategy also reflects a broader view of building ecosystems where LLM technologies are central in enhancing product offerings and operational efficiencies. This is evident in NVIDIA’s foray into the LLM domain with “NeMo Megatron,” designed to optimize GPUs for processing massive datasets, thus supporting the infrastructure needs of large-scale LLMs.

Challenges for startups in this scenario

At this point, you already understand how immense the barriers to entry in the LLM market are, particularly due to the enormous scale of data and computational resources required to develop and train these models. With their deep pockets and extensive infrastructure, big tech companies have a significant advantage in this space. They possess vast data repositories harvested from their wide array of services and platforms, which are critical for training sophisticated LLMs. Moreover, the computational power needed to process this data and train the models is substantial, often involving state-of-the-art GPUs and extensive cloud computing resources — which, again, only the Magnificent 7 have. This creates a high entry threshold difficult for startups to meet without significant investment or innovative approaches to model training and data utilization.

Competing with the brand and reach of big tech adds another layer of challenge for startups in the LLM space. Big tech companies not only have the advantage of brand recognition but also established distribution channels that can seamlessly integrate these LLM technologies into existing products and services, thereby enhancing their value proposition and user engagement. For a startup, building a reputation and gaining traction in a market where consumers and businesses are accustomed to the reliability and sophistication of products offered by big tech giants is a daunting task. This scenario requires that startups not only innovate technologically but also find unique market niches or develop disruptive business models that can offer compelling alternatives to the offerings of their much larger competitors.

It is not a coincidence that the once LLM startups such as Anthropic and OpenAI are accepting those investments. Without them, it’s extremely difficult to scale and continue to improve their own models.

So, is there anything left for startups here?

The combined effect of these barriers means that for startups to carve out a space in the LLM market, they must leverage agility, focus on niche applications where they can offer distinctive value, or pursue partnerships and collaborations to amplify their reach and resources. Innovation in model efficiency, data utilization, and deployment strategies also plays a crucial role in overcoming these challenges and making a mark in the competitive landscape dominated by big tech.

For instance, small language models (SLMs) present a promising frontier. Unlike their larger counterparts, SLMs require significantly less data and computational resources to train, making them more accessible for startups. These models can be customized for specific industries or tasks, such as legal document analysis or niche language translation services, where big tech’s one-size-fits-all approach may not be as effective.

The agility and innovative capacity of startups allow them to adapt to market changes and technological advancements quickly, a crucial advantage in the rapidly evolving AI landscape. By focusing on areas underserved by big tech, such as localized AI solutions for small businesses or specialized educational tools that leverage LLMs for personalized learning experiences, startups can carve out significant market segments. Furthermore, the regulatory environment and ecosystem support play pivotal roles. For example, the EU’s AI Act could level the playing field by imposing standards that ensure transparency and fairness, benefiting nimble players who can adapt quickly.

Several startups have already demonstrated success in leveraging these opportunities. For example, companies like Cohere and AI21 Labs have made strides in offering AI as a service, focusing on natural language understanding and generation with a more accessible and flexible platform than larger entities' offerings. Additionally, startups like Hugging Face have emphasized community and open-source collaboration, rapidly becoming central hubs for the latest AI models and techniques. These examples underscore the potential for startups not only to compete but also to lead in specific sectors of the LLM market by focusing on innovation, specialization, and leveraging regulatory changes and ecosystem dynamics to their advantage.

But will Cohere, AI21 Labs, and Hugging Face be the next top of the S&P500? Not really, as Google, Nvidia, and Amazon also invested in them. So, at least in the current state of the market, there seems to be no escape from the dominance of the big techs.