The End of the Beginning?
Layoffs, declining traffic, and the first lull in AI's current hype cycle
Hope everybody is having a great summer! Thank you to our now 1,200+ subscribers, we appreciate all the continued support! Please continue to share with colleagues and friends :)
It’s hard to believe ChatGPT was released less than eight months ago (the official launch date was November 30, 2022); in that short period, we have witnessed a new renaissance in AI. Hundreds of new startups have launched, thousands of products released, and millions of dollars in investment capital have poured into the space. Even public companies like Microsoft and NVIDIA are seeing their stocks hit all-time highs driven largely by their moves around AI.
Eight months in, however, we are now starting to see the first signs of that hype subsiding:
Earlier this week, The Information published a piece detailing how two high-flying Generative AI startups - Jasper and Mutiny - are laying off some workers, “striking a downbeat note amid a wave of investor euphoria for the sector.”
The week prior, news came out that ChatGPT saw its monthly traffic in website visitors and users decline in June for the first time, leading a Similarweb analyst to observe that “decreasing traffic is a sign of the chatbot's novelty wearing off”.
And to top it off, the maker of ChatGPT (and one of the most prominent names in AI) OpenAI is now being investigated by the FTC for possibly violating consumer protection laws.
All of this could lead one to think the promises of AI are falling flat.
We don’t believe that. Instead, we are likely just experiencing the first lull since the current AI hype cycle kicked off last fall. Teams are still building incredible products, the open-source community is thriving, and real value is being generated thanks to AI (and now even the world’s largest company is getting in the game).
It is still helpful to dissect what exactly is behind these negative headlines, and what we can learn as a community to continue innovating.
In other words - let’s stop freaking out, and peel back what’s actually going on here.
So Why The Distress?
Let’s take each of these recent news events and go one layer deeper:
Layoffs at Generative AI Startups
We should start by stating the obvious: building and scaling startups is INCREDIBLY hard. Not only is it difficult to build a startup in normal boom times, but it’s another order of magnitude more difficult in a down market.
Just look at the stats on layoffs.fyi; thousands of tech workers have been laid off since the beginning of the year, ranging from small startups to giants like Microsoft and Meta. The current success of a business does not necessitate immunity from workforce reductions. More than anything, high-growth tech companies overhired during the ZIRP period, and that phenomenon has been slowly unwinding.
Jasper rocketed out of the gates in 2022, growing from ~150 employees in Jan’22 to nearly 600 today per LinkedIn. Mutiny also experienced rapid growth, upping headcount by ~300% over the past two years. So from that perspective, it’s not terribly surprising that both have made decisions to shed some of the team (Jasper CEO Dave Rogenmoser’s note on the layoff is worth reading).
However, it is important to recognize what can happen to first movers in a rapidly evolving space. When Jasper first launched, they were one of the only business applications powered by GPT, and as such was often the first use case for AI that most business users experienced. In the months since, several new AI-powered copyrighting apps have emerged, including Copy.ai, Writer, Writesonic, Copysmith, etc…not to mention popular general-purpose language tools like ChatGPT and Bard. As with any other category, first-mover advantages rarely last, and the only way to stay ahead of the competition is to continually evolve your product, adapt to what users want, and establish long-term moats.
ChatGPT Slows Down
Earlier this month, data from analytics firm Similarweb showed that ChatGPT’s monthly traffic declined for the first time since its November 2022 launch:
Worldwide desktop and mobile web traffic to the ChatGPT website, chat.openai.com, dropped 9.7% from May to June, according to preliminary estimates. In the U.S., the month-over-month decline was 10.3%.
Worldwide unique visitors to ChatGPT’s website dropped 5.7%. The amount of time visitors spent on the website was also down 8.5%.
ChatGPT still attracts more worldwide visitors than bing.com, Microsoft’s search engine, or Character.AI, the second most popular stand-alone AI chatbot site. Worldwide visits to Character.AI dropped 32% month-over-month, although traffic is still up tremendously from where it was in June 2022, when the company founded by ex-Google engineers was just getting started.
It’s clearly noteworthy that the explosive growth of ChatGPT has reversed course. But there are likely some other factors at play here:
Schools’s out → Students, who comprise a significant chunk of ChatGPT users, are in their summer holidays and likely using ChatGPT much less than when they were in school. So this may mean that the traffic decline is just a temporary blip until the holidays end. We can test this hypothesis to see if traffic picks back up in September…
Direct vs. Indirect → The unprecedented popularity of ChatGPT served as a catalyst for Google to wake up and push out their own chatbot Bard in February. Also, Microsoft is now exposing ChatGPT into Bing and their own Office products through their partnership with OpenAI. So while direct traffic to ChatGPT may be declining, usage of OpenAI’s products in totality may actually be INCREASING.
We will see if this decline is temporary or more permanent, but it’s an important reminder that RETENTION matters. A good recent example in another space is Threads, Meta’s alternative to Twitter. Threads dethroned ChatGPT as the fastest-growing consumer application in history, reaching 100M users in a mere five days. However, Similarweb reported that retention is fleeting, with just DAUs down by half (~50M to ~24M) just a week later.
What does all of this mean? The number of signups for every new AI product will only continue to get larger faster. But what really matters is if the product can deliver enough value to keep users ENGAGED enough to stay on the platform.
FTC Investigation
Last week, it was reported that the Federal Trade Commission was investigating OpenAI for “possible violations of consumer protection law, seeking extensive records from the maker of ChatGPT about its handling of personal data, its potential to give users inaccurate information, and its ‘risks of harm to consumers, including reputational harm.”
We view this investigation less as an existential threat to OpenAI, but rather an indication that AI has reached a level of importance for regulators akin to the rest of BigTech. In the past year, regulators have played a much more active role in the tech landscape. The FTC lost its effort to block Activision’s $69B sale to Microsoft, while the DOJ is reportedly planning an antitrust lawsuit to block Adobe’s $20B acquisition of Figma.
If anything, startups and scalers should recognize that with AI clearly in the mainstream of public consciousness, they may be in the regulatory crosshairs far earlier than the norm.
What’s the Counter Argument?
Every hype cycle goes through a form of rapid adoption curve: an initial euphoria period of overinflated expectations followed by a reversion to the mean. While we are in the very early days of this particular technological shift, we are also seeing many green shoots, which typically aren’t as common this early on. Companies such as MosaicML (gen-native enabler) and Casetext (gen-enhanced application) have been acquired for eye-popping prices, $1.3B and $650M, respectively. M&A at these levels is very rare this early in a cycle, so already this platform shift has delivered real exits (not to mention market movement) over just paper markups.
Also, unlike many other hype cycles (e.g., Crypto), we are not just seeing early adopters or a small portion of the public use these products. Within the technology sector, we see the largest companies like Apple, Microsoft, Amazon, Google, Meta, Oracle, Adobe, and Salesforce all making major announcements around new investments into their own LLMs or new product features leveraging models.
AI isn’t just limited to the tech sector either; the vastness of the use cases is exemplified by how many NON tech companies have already developed ways to use AI internally:
In retail, Coca-Cola is leveraging AI for creative efforts
In banking, Goldman Sachs built a ChatGPT-style AI in-house to assist developers with writing code
In food and beverage, KFC & Taco Bell use AI to lower food costs and better forecast weekly inventory supply
We are clearly still in the early innings here!
What Can We Learn From All This?
Moats matter. This is the first major technological shift that we’re seeing where companies can leverage the technology via an API call. In previous shifts (e.g., Cloud Computing) there needed to be physical operating system changes that were required (e.g., moving from on-prem > cloud). As a result, it is much easier for newer incumbents to emerge and build using underlying FM technology. As an early-stage company, you cannot just be a thin wrapper on another infrastructure provider.
Competition is coming. Every company, small or large, is recognizing that many dollars will be up for grabs in this major platform shift. Every space will quickly fill with new competitors. Be aware of the advantages that BigTech and incumbents have (e.g., access to large pools of capital, distribution, network effects) and also recognize that many newer entrants that are building using the newest technologies which may be 10x cheaper (e.g., open source).
Users are fickle. Many GenAI apps today are prosumer apps and driven through product-led-growth (PLG). Consumers are always looking for the coolest products which means they have no problem jumping from product to product and can do so easily at no cost to them. Many of these early-stage prosumer apps won’t be as sticky as B2B contracts of the prior SaaS era, so driving user retention will be doubly critical for company success.
Use capital wisely. It is easy to burn millions in this space trying to find product market fit or trying to build out the best model in the foundation model. Companies may spend more than expected building, training, and fine-tuning models or running hundreds of queries to achieve the best output for the end-user. While VCs may be investing behind many ideas in the AI/ML space, just remember that capital is not always abundant.
Unicorns will emerge, but not every company will be one. There are likely going to be many companies that emerge that are successful, but may not scale to $1B+, and that’s OK! In fact, we believe that in this new era of AI, we will see many smaller companies (<25 FTEs) being built that still drive significant value.
Keep building! Every prior era had an early set of companies that rocketed to stardom and eventually fizzled out. Think of Webvan and Pets.com in the early days of the Internet, or Fruit Ninja and FourSquare in the early days of mobile. In the end, billions of value was generated in both of these waves. So, while many of the early adopters of companies building in the AI wave may end up disappearing, there WILL be winners. That is simply part of the game.
Funding News
Below we highlight select private funding announcements across the Intelligent Applications sector. These deals include private Intelligent Application companies who have raised in the last two weeks, are HQ’d in the U.S. or Canada, and have raised a Seed - Series E round.
New Deal Announcements - 07/06/2023 - 07/20/2023:
