#438: AI Assistant Search Could Upset Google's Moat, & More
1. AI Assistant Search Could Upset Google's Moat
Last week, OpenAI upgraded1 ChatGPT’s search capabilities, as rumors swirled2 that Meta is building an in-house search engine to capitalize on its AI’s capabilities. Like Perplexity, OpenAI and Meta are offering AI-driven assistants that collectively could create a significant challenge to Google’s dominance in search. In traditional search, users must comb through a list of hyperlinks ranked by relevance, while the new AI-powered assistants respond with coherent, conversational narratives that they have synthesized from various information sources.
The shift from traditional search to AI-first workflows could upgrade the user experience significantly by saving time and improving the quality of results. Google’s link-based advertising model could be disrupted in the process if, in response, traditional search ad inventory declines. In a new model, AI-first workflows could integrate ads seamlessly and pleasingly into interactions between users and AI assistants.
Importantly, the evolution toward AI-first information discovery could transform search from a “product” to a widely available “feature,” potentially eroding Google’s moat. If users were to gravitate toward the AI assistants that Perplexity, OpenAI, Meta, and even Google’s own Gemini now offer, search could shift from a “product” that users harness, to a “feature” that AI assistants use in answers to user questions or requests. In this new paradigm, the companies hosting AI assistants will choose which search engine to use, as well as when and how to use them. Perhaps those companies will vertically integrate, creating their own search engines and taking share from Google and Bing.
2. Waymo Funds At A Lofty Valuation While Sending Mixed Signals About LiDAR
Last week, Waymo raised $5.6 billion, boosting its valuation to $45 billion—approximately 580x its annualized revenue, by ARK’s estimate.3 For perspective, as of November 4, 2024, Tesla trades at ~7x forward revenue, and OpenAI is valued at ~13x forward revenue.4
Last week, Waymo also published a paper on “EMMA”—its End-to-End Multimodal Model for Autonomous Driving. Interestingly, EMMA uses data only from cameras, not radar and LiDAR. Apparently, the low volume of LiDAR- and radar-derived data has limited LiDAR 3D sensing encoders, making them less generalizable and sophisticated than camera encoders for the purposes of this research.
Powered by a camera-based perception system, an end-to-end model would seem to be emulating Tesla’s approach to autonomous driving models. That said, according to Waymo’s paper, EMMA will face challenges in real-world deployment without LiDAR. In a recent podcast, Waymo’s CEO also said that, while it could enable driver assistance, a camera-based system would not meet Waymo’s performance standards.5 In other words, while the paper hinted at abandoning LiDAR, Waymo seems committed to it for the time-being.
In our view, even if LiDAR were less than ~$5,000 per unit today, it is adding costs and complexity that could limit the scalability of Waymo’s autonomous vehicle fleet.6 Tesla is aiming to price its Cybercab below $30,000, sensors included.7 Waymo seems to believe that incorporating LiDAR will add to safety and be worth the
3. Cures For Disease Could Lead To Better Biotech Business Models
Biotech companies are using new molecular tools, notably CRISPR gene editing, to help cure the rare diseases that millions of people suffer. While the promise for patients is obvious, many analysts and investors seem to doubt the investment merits of curing rare diseases. The market is large, as demonstrated below, but many doubt the business economics.
Developed by CRISPR Therapeutics, Casgevy is a cure for sickle cell disease that probably generated its first meaningful revenue in the third quarter, to be reported this week. Analysts do not seem sure how to model the commercialization glidepath for Casgevy and other such therapies. As a result, investors seem to be concerned that a cure—while good for humanity—will not be good for business: monthly dosing and annuity-like revenues seem superior to a one-and-done cure. They also seem disinterested in rare diseases, perhaps thinking that they are unlikely to lead to blockbuster drugs.
According to ARK’s research, those conclusions underestimate the profound business opportunities that curing disease could create, as shown in the table below. Calculated for the average rare disease, a cure should be more than twice as valuable as a competing precision drug dosed regularly that abates the symptoms of the underlying condition. Yet, precision therapies do not exist for most rare conditions. For most rare diseases, therefore, cures should prove 20x more valuable than the current prescription drugs on offer.
Cures are likely to result in better business models for three reasons:
- Reimbursement for cures is based on avoiding a lifetime worth of medicine, medical, and human health costs.
- Cures frontload cash flows in a drug’s commercial life cycle, giving companies quicker access to capital either to reinvest in research and development or to return to shareholders.
- Cures capture the lifetime value of each patient upon the first dose, side-stepping the competitive intensity associated with patent expiration and me-too drugs.
The graph below compares the lifetime revenue curve for a cure addressing the average rare disease to that for a precision therapy with traditional dosing.
As suggested in this model, early in its lifecycle, the cure generates 15x more revenue than the precision therapy by capturing the lifetime value of each incremental patient dosed. By contrast, a traditional treatment accrues revenue on a shallower but smoother trajectory based on the cumulative number of patients under treatment. Because cures aim to eliminate the market they address, later stage revenues fall off more quickly than those for traditional treatments. That said, the present value of a cure should be higher than that of the periodic dosing regimen. In this model, the cash flow of the cure is worth more than two times that of the precision therapy: $2.1 billion versus $900 million.8
More prevalent and more severe rare diseases could prove substantially more valuable. A salient example is Hereditary Angioedema (HAE). This genetic disease affects 7,000 patients in the US, reduces quality-adjusted lifespans by roughly 20%, and costs $500,000 per year for medication to manage it.9 Two weeks ago, Intellia disclosed Phase 2 results for a gene-editing-based prospective cure for HAE. The results suggested that a one-time infusion cured 73% of the 11 patients, likely on a permanent basis, but the stock traded off roughly 20% on the release.10
ARK estimates that a cure for 70% of HAE patients would command a price of $2.5 million per dose.11 Based on just the US patient population, the addressable market for this cure is $17 billion, and the global market likely double that amount. The present value of a cure for HAE that penetrates half the global market over five years could total $10 billion.12 At an equity market cap of less than $1.5 billion and with other meaningful assets in its pipeline, Intellia’s current valuation suggests that analysts and investors do not believe this analysis.
To be sure, Intellia’s Phase 3 results may not be as promising as its Phase 1 and Phase 2 data, and precision therapies with traditional dosing profiles also are targeting the HAE market. Also likely, however, Intellia could re-dose patients who did not respond to the initial treatment, possibly increasing the overall cure rate and prospective market size. Also possible, cures could penetrate patient populations more quickly than traditional treatments, given their superior results.
Though the market seems to misunderstand the prospective value of curing disease, ARK’s work suggests that cures could become extremely valuable assets. Measured like-for-like, they appear to be 2.3x more valuable than traditionally dosed precision therapies. The first of these cures to hit the market—Casgevy—should generate revenue to be disclosed in its third-quarter earnings report this week. Longer term, in our view, the financial results associated with cures will be as difficult to ignore as the profound impact they will have on human health.
-
1
OpenAI. 2024. “Introducing ChatGPT search.”
-
2
Reuters. 2024. “Meta builds AI search engine to cut Google, Bing reliance, the Information reports.”
-
3
According to Waymo, the company completes more than 150,000 paid rides per week. The assumptions of our estimate are as follows: each ride is ~5 miles and priced at $2 per mile, roughly in line with Uber today. See Roof, K. et al. 2024. “Alphabet’s Waymo Valued Above $45 Billion After Funding.” Bloomberg. See also Waymo. 2024. “Waymo One is now providing over 150,000 paid trips.” X.
-
4
Based on data from CapIQ data for Total Enterprise Value / Total Revenue Next 12 Months retrieved 11/4/24; and Downing, F.. 2024. “OpenAI Investment Thesis.” ARK Investment Management LLC. The OpenAI multiple is based on the latest valuation round and calculated as of 10/15/24.
-
5
See No Priors. 2024. “Ep. 87 | With Co-CEO of Waymo Dmitri Dolgov.” YouTube.
-
6
Corrall, C. 2024. “Tesla’s Cybercab robotaxi is here — and it could cost less than $30K.” TechCrunch.
-
7
Ibid.
-
8
This model assumes the same patient penetration rates, a 10% discount rate, and cure and precision therapy pricing consistent with the methodology used by The Institute for Clinical and Economic Review (ICER), which assumed $100k per quality-adjusted life-year.
-
9
Chaguturu, S. “Number of HAE treatments increasing.” CVS Caremark.
-
10
Cohn, DM. et al. 2024. “CRISPR-Based Therapy for Hereditary Angioedema.” N Engl J Med.
-
11
The fair price for existing prophylactic treatments is roughly $250,000. At $100,000 per life year, patients still lose out on $20,000 in quality of life annually. The present value of $270,000 over the patient’s 23 years of life at a 5% discount rate is $3.4 million. Adjusted to $2.5 million for the 30% of patients who still have symptoms (albeit reduced).
-
12
Assumes 50% patient penetration by year 5, a 20% tax rate, and 60% margin, all at a 10% discount rate.