Multiple developments in the heated race to the Enterprise AI leaderboard: vertical vs horizontal
The release of OpenAI’s Strawberry agents is set to become more widely accessible, a significant advantage for vertical AI development. A recent example is the success of Klarna’s AI initiatives.
Simon Taylor from Fintech Brainfood wrote an insightful piece on how competition in Fintech AI is evolving. He suggests that companies like Stripe, Ramp, and Klarna—members of the Launchbay25 index—stand to benefit from the new opportunities in AI agent development.
A fresh example from the SMB fintech space is Gusto, another Launchbay25 member, which has just opened a waitlist for its AI assistant tailored for small business owners.
Glean, high liquidity company on Launchbay platform, recently closed a new funding round and released a report highlighting its impressive revenue growth and adoption. The results reflect strong investor confidence in the development of horizontal enterprise AI, further strengthening SaaS incumbents like Salesforce through seamless integrations.
Glean reported a threefold increase in revenue for 2023, aligning with Microsoft Copilot's stated growth rate of "60% per quarter," as noted by Satya Nadella.
New day for everyone focused on inference
The release of OpenAI's new model, "Strawberry," is also having a major impact on the semiconductor industry.
Doug O'Laughlin from the Fabricated Knowledge newsletter explains: “What does this mean for semiconductors? A huge increase in computing demand, but this time it’s focused on inference, not just training.” He highlights how the new scaling law emphasizes inference-time reasoning, or the "Chain of Thought."
In essence, OpenAI's o1 model and its focus on inference scaling are poised to reshape the business landscape for AI hardware companies like Groq, Cerebras, and SambaNova. As o1 enhances reasoning abilities at inference time, the demand for specialized hardware is expected to rise, benefiting these companies.
Groq, with its Language Processing Units (LPUs), stands to gain from this trend. Its architecture, designed for sequential processing and high memory bandwidth, aligns perfectly with the needs of inference-intensive models like o1, potentially driving broader adoption of its technology.
Cerebras’ Wafer-Scale Engine (WSE-3), which tackles memory bandwidth constraints crucial for efficient inference in large models, is also well-positioned. The shift toward inference-time scaling supports Cerebras' strategy, reinforcing its competitive standing.
SambaNova, having recently launched its cloud inference platform with record-setting performance on models like Llama 3.1 405B, is similarly poised to capitalize on the growing emphasis on inference speed and performance.
While less talked about than AI chip prices, advancements in nuclear technology are equally intriguing. Driven by rising power demands, Oracle plans to utilize mini nuclear reactors to power its AI operations, and Microsoft is exploring a "nuclear strategy" for its energy needs.
New AI valuations
OpenAI's upcoming financing round is anticipated to value the company at $150 billion, translating to a share price of nearly $360. While there are fresh asks at this price, secondary market buyers have yet to fully support the new valuation, with bids currently ranging between $185 and $200 per share.
We asked Strawberry to analyze whether this valuation is reasonable, based on the open data we provided. Its response was: “Evaluating whether this valuation is reasonable ultimately depends on future developments in OpenAI's financial performance, technological advancements, and the broader AI industry's trajectory.”
This valuation is undoubtedly influencing the ambitious valuations of other liquid AI startups. A recent example is CoreWeave, which is in talks for a share sale at a $23 billion valuation, marking a 21% increase in just three months.