Gen AI's performance review-- compatible with life or failure to thrive?

Profuse amounts of capital have been thrown into the AI pot in the past two years, and the fog of frenzy is slowly clearing. Most of the funding for generative AI (gen AI) has been directed towards the infrastructural level due to several key factors that prioritize foundational technologies over end-user products. As we’re starting to see the first hype wave of Gen AI subside, the infrastructure companies have emerged as the largest venture backed class of AI startups for several reasons:

1. Infrastructure and Computational Resources: Training large-scale generative models requires substantial computational resources, including high-performance computing clusters, specialized hardware accelerators (e.g., GPUs, TPUs), and cloud-based infrastructure. While compute costs had seemed to become negligible pre ChatGPT era, the consumption needs of large language models have temporarily spiked up cost per run. Promising deliverables that require high capex have generally pushed startups to raise large rounds of subsequent funding as they scale in ability and compute needs.?

2. Platform Development and Tooling: Platforms and tools for developing large language models, such as Anthropic, Together AI, and Hugging Face's Transformers library, requires investment in software engineering, user interface design, and developer support. Funding at the infrastructural level supports the development and maintenance of these platforms and tools, which provide developers with the necessary infrastructure and resources to build and deploy generative AI applications. These platforms will continue to be important as developers building closer to end consumer may have more complex observability or tooling needs.?

3. Open-Source ethos: Much of the progress in gen AI has been driven by open-source and collaborative development efforts, where researchers and developers contribute code, models, and datasets to the community. Funding at the infrastructural level supports these collaborative initiatives by providing resources for community engagement, project management, and infrastructure maintenance. Much of these communities have reignited interest in edge or localized compute, as well as driven frameworks for smaller more vertical focused models that are less expensive to train.?

Thus, while the number of application classes of Gen AI startups (Gen AI for x) seem to rapidly rise, the majority of the funding in this space as of the end of 2023 has gone to infrastructure startups. However, while we see independent startups continue to balloon in size and capabilities, large incumbents stand to have a great built-in advantage and have also been big parts of funding these large rounds.?

Fighting for space alongside incumbents

Drawing parallels to other platform shifts of the past such as mobile and cloud, In the tech world, incumbents like Google, Amazon, Apple, and Microsoft have a history of dominating market shifts by building expansive ecosystems around emerging platforms. From the early days of the internet to the rise of mobile and cloud computing, these giants have consistently captured significant value by leveraging their vast resources, infrastructure, and user bases.

Each platform shift has presented incumbents with an opportunity to extend their reach and influence, and they've capitalized on it brilliantly. For instance, during the internet era, companies like Google capitalized on search and advertising, while Amazon transformed e-commerce and cloud computing. Similarly, Apple revolutionized personal computing and entertainment with its lineup of iPhones, iPads, and iTunes, while Microsoft dominated the software market with Windows and Office. They have all consistently captured value from each of these paradigm shifts (which is certainly correlated with stock sentiment/performance though does not sufficiently account for all of it).?

Now, with the advent of AI, we're witnessing a new platform shift, and once again, the incumbents are poised to capture value with their ever-expanding ecosystems. Companies like Microsoft, with large investment in OpenAI, NVIDIA with its diversified strategy of investing in different verticalized AI early stage startups, have already established strong footholds in the AI space.?

So, where does this leave startups? Unlike previous platform shifts where startups could compete by building complementary products or services within the incumbents' ecosystems, the AI era requires a different approach. Overall, while the first wave of Gen AI funding has been spent in majority at the infrastructure level and early innings of the application layer, as use cases continue to mature, we can expect to see the fruits of these investments manifest in the stepchanges of innovative gen AI products and subsequent opportunities.

The natural progression has landed in attention closer to end users in enterprise, where Saas was once the darling of entry into vertical plays, the excitement and shifting acceptance in enterprise towards AI has played a role in enthusiasm.? There has been a growing emphasis on developing specialized AI infrastructure for specific use cases and industries. For example, in healthcare, startups and established companies are building AI platforms tailored to medical imaging analysis, patient diagnosis, drug discovery, and personalized medicine. Similarly, in finance, AI infrastructure is being developed to support algorithmic trading, risk assessment, fraud detection, and customer service automation.

However, as the AI landscape continues to mature, there's a new push forward towards innovation in reducing human to algorithm interfacing lag time and friction.? Outside of the example of building generalized intelligence to bridge this gap, other motivations include the product design capability of being able to "speak" a product to be. Unlike traditional AI applications that primarily focus on data analysis, pattern recognition, and decision-making, generative AI aims has quickly proven capability in generating baseline content, such as text, images, audio, video and code, that is human-like. As we push the bounds of more complex tasks,such as architecting and writing whole units of products from a spec,? it can revolutionize our speed of creation, efficiency and productivity.

Moving forward, we can expect to see increased innovation in generative AI products that enable natural language interaction, storytelling, content generation, and creative expression. Startups and companies are likely to leapfrog off of current generations of tools, platforms, and applications that leverage generative AI to enhance human-machine interaction, automate content creation, and personalize user experiences. We at Tau are excited about the productized layer of Gen AI, which may soon come!

Primary author of this article is Sharon Huang . Originally published on “Data Driven Investor .” ? These are purposely short articles focused on practical insights (we call it gl;dr — good length; did read). See here for other such articles. If this article had useful insights for you, comment away and/or give a like on the article and on the Tau Ventures’ LinkedIn page , with due thanks for supporting our work. All opinions expressed here are from the author(s).