ARK • Disrupt

Last week, on the heels of its GPT4 release, OpenAI announced a product extension for ChatGPT allowing the chatbot to interact with external data and services. With plugins, ChatGPT now can search for real-time information on the web, order groceries from DoorDash, and book flights, hotels, or rental cars on Kayak. In short, plugins have transformed ChatGPT into an app platform that could challenge the iPhone.

In our view, this innovation will expand the economic potential of ChatGPT and pose a competitive threat to both Google search and Apple’s App Store monopoly. Google’s revenue depends on consumers and businesses searching and moving from website to website. With plugins, ChatGPT inverts the process: consumers and businesses stick with ChatGPT as it discovers and presents options. Why bounce from site to site when a single resource can respond in natural language to the nuances of vacation plans with price comparisons, book dinner reservations, and surface that perfect out-of-the-way beach condo?

Well and good for the vacationing consumer—but not so good for Google search traffic. Measured against ChatGPT and Bing, Google already is bleeding traffic share, as shown below. With plugins, that trend seems likely to continue, if not accelerate.

ChatGPT could threaten Apple’s App Store. Who needs a special purpose app when a general-purpose interface can provide answers across the open web? In other words, the language model becomes the operating system, giving companies a way to avoid Apple’s 15-30% platform tax.

In our view, tech’s center of gravity is shifting dramatically. Language models are performing super-exponentially, leaving consumer-facing mega-cap tech companies flat-footed. Assuming aggressive investment and cost declines averaging ~3X per year, we could see the capability of artificial intelligence (AI)¹ systems increase two-hundred-fold over the next three years, as shown below. This progress would be equivalent to the gain in iPhone performance over the 16 years since it launched in 2007.

[1] We define “AI capability” as the aggregate forecast amount of capital invested in training AI systems adjusted for increasing AI performance per dollar invested due to cost declines.

Last week, short seller Hindenburg Research published a research report on Block (SQ) claiming that Cash App, Block's digital wallet, has inflated user numbers and could lose significant revenue as regulators change measurement rules. In response to a 15% drop in its shares on Thursday, we and the investor community, on balance, defended Block against wildly misleading claims that included few credible data points.

Exemplifying its questionable claims, Hindenburg cited a Pandemic Unemployment Assistance (PUA) fraud case and provided a cropped screenshot of an indictment, stating that "the only peer-to-peer electronic payment processor mentioned was Cash App." What Hindenburg failed to report was that the criminal rapper fraudulently accessed more than $1,200,000 with electronic benefit transfer (EBT) debit cards issued by Bank of America and cashed them out largely in the traditional banking system—ATMs, bank branches, debit cards—not fintech apps. In fact, the rapper used money transfer/fund transfer services, including Cash App, to access only $85,130 of the fraudulent funds, compared to $375,177 from ATMs, $39,180 from bank branches, and $205,273 from debit cards issued by traditional banks like Bank of America.

Even if Hindenburg’s claims prove unfounded, we do wonder how they will impact Block and the broader fintech ecosystem. Regulators have been focused on combating peer-to-peer payment fraud on Venmo, Zelle, PayPal, and Cash App and might intensify those efforts, perhaps forcing companies like PayPal and Block to report unique users instead of active accounts. That said, we believe full-scale regulatory intervention is unlikely given the current banking crisis, including the focus on cryptocurrencies.

Last week, Coinbase received a Wells Notice from the Securities Exchange Commission (SEC), which often signals the high probability of enforcement actions. In response, Coinbase suggested that the Wells Notice could be related to its retail and institutional trading markets, staking services, and self-custody wallet—Coinbase Wallet. While Coinbase received its first Wells Notice last year as it was preparing to launch Lend, this Wells Notice is much broader in scope and atypically vague, with few specifics about products purportedly breaching securities laws. Coinbase has signaled its readiness to challenge the SEC’s enforcement recommendation, seemingly willing to escalate the issue to the Supreme Court during the next few years. Cases like Grayscale, Binance.us, and Ripple, suggest that the US legal system will protect the due process rights of crypto firms and require the SEC to substantiate its allegations, hopefully leading to regulatory clarity.

Coinbase maintains that, after reviewing its listing criteria with regulators and receiving no negative feedback, it has blocked 90% of token applicants from its exchange. In fact, it had listed many of the tokens on its exchange before its debut as a publicly traded company in 2021. Moreover, in the S-1 it submitted to the SEC, a necessary step before going public, Coinbase referenced its staking service, which now seems to be a problem for the SEC, 57 times.

In our view, Coinbase has a compelling case that should result in legal clarity for the crypto industry. Importantly, its direct integration with decentralized public blockchain infrastructure—including its Ethereum scaling solution, Base, and products like USDC and Coinbase Wallet—should help Coinbase scale globally, lessening its dependence on US regulators.