Why most AI startups in the bubble are doomed to fail
According to the available knowledge, most the current wave of AI startups seem to piggy back on using APIs or data intelligence provided by large language models (LLMs) common datasets created by OpenAI, Anthropic, Google, Microsoft and a few others.
All big tech companies, as well as big corporations (say SAP) are in a race building these models in their specific domains.
Numbers wise, the AI startup space is freaking booming, with an estimated 30,000 AI startups worldwide. The total funding for AI startups reached $53.2 billion in 2022, up from $37.5 billion in 2021. This growth is being driven by the increasing adoption of AI across a wide range of industries, including healthcare, finance, and transportation.
Despite the rapid growth, the AI startup space is still relatively young and there are a number of challenges that AI startups face, including:
- Technical challenges: Developing AI applications is complex and requires a high level of technical expertise.
- Data challenges: AI applications often require large amounts of high-quality data to train.
- Talent challenges: There is a shortage of skilled AI talent.
- Funding challenges: AI startups often require significant funding to develop and commercialize their products and services.
The key problem for the AI startups where the data intelligence comes from. For now, it looks like almost all of it (not sure the real percentage, but probably 80% or more given the old numbers) comes from the open web, training on freely available data. And very little comes from other data points such as books and specific databases. ChatGPT is rumored to be trained on some specific data like Amazon product reviews etc.
See the problem? This whole bubble comes from LLMs which are based on limited open web dataset.
Yes, If you can create a project on the weekend, a competitor also can create a similar project likewise.
Indeed, there may be a coding genius involved, a 10x programmer, inspiring UI. It can take several weeks for others to replicate. But they can be replicated. If you don’t have a special dataset or user case. It’s gonna be a bloodbath.
If you have a specific proprietary dataset which no one else has, this can’t be easily replicated by any LLM model you are on the winning party. Still If you built anything on generic datasets which goes viral, any big tech company can replicate it in-house at a scale very easily.
Sounds familiar? well my assumption is most of these AI venture funds will be wasted just like in the dot com boom evaporated most of the wealth it created. Remember?
Dot Com Boom in the 2000s
The dot-com bubble was a period of rapid growth in the internet sector that began in the mid-1990s and collapsed in 2000. The collapse of the bubble resulted in the loss of trillions of dollars in investor wealth, with estimates suggesting that the total losses could be as high as $5 trillion.
The exact percentage of projected value creation that was lost during the dot-com bust is difficult to determine due to the varying estimates of total losses and the subjective nature of projected value creation. However, some estimates suggest that the losses could be as high as 80% of projected value creation.
For example, a 2003 study by McKinsey & Company estimated that the dot-com bust destroyed $5 trillion in shareholder wealth. This figure represents about 80% of the $6.3 trillion that was invested in internet companies during the bubble period.
Based on Available Google Data
The dot-com bubble was caused by a number of factors, including:
- Speculative investment: Many investors poured money into internet companies without fully understanding their business models or potential for profitability.
- Overvaluation: Internet companies were often valued at multiples of their actual earnings, based on the expectation that they would continue to grow rapidly.
- Lack of regulation: The internet was largely unregulated at the time, which allowed companies to make exaggerated claims about their products and services.
The collapse of the dot-com bubble had a significant impact on the economy. It led to widespread job losses in the technology sector and a decline in consumer spending. The bubble also caused a loss of confidence in the stock market, which led to a further decline in stock prices.
Not on the same AI startup footing but dot come bubble was kinda created on a technology promise which evaporated soon the industry learned that the promised value creation lacks the actual technology backing. Actually what promised as dot com boom came 7 years later as the mobile revolution with the legendary Steve Jobs introduction of the iPhone, with full stack development ability on a, gps, camera, cellular, apps and sensor laden package as a mobile.
The dot-com bubble was a major economic event that had a lasting impact on the internet industry and the economy as a whole.
Similarities between AI startup space and dot-com boom.
- High levels of investment: Both the AI startup space and the dot-com boom saw high levels of investment from venture capitalists and other investors. This influx of cash can lead to companies with unsustainable valuations and unrealistic growth expectations.
- Speculative investment: Many investors are pouring money into AI startups without fully understanding the underlying technology or the potential for profitability. This speculative investment can lead to a bubble where stock prices are inflated beyond their true value.
- Overvaluation: AI companies are often valued at multiples of their actual earnings, based on the expectation that they will continue to grow rapidly. This overvaluation can make it difficult for companies to justify their valuations and can lead to a correction if growth expectations are not met.
- Lack of regulation: The AI industry is still relatively unregulated, which allows companies to make exaggerated claims about their products and services. This lack of regulation can lead to investor confusion and can make it more difficult to identify which companies are truly innovative and which are simply hype.
These similarities suggest that the AI startup space is at risk of a bubble burst. However, it is important to note that there are also some key differences between the two periods. For example, the AI industry is much more mature than the internet industry was in the late 1990s. Additionally, AI is a more transformative technology than the internet, and it has the potential to solve some of the world’s most pressing problems.
As a result, it is possible that the AI startup space will avoid a bubble burst. However, investors should be aware of the risks and should carefully evaluate the companies they invest in.
Here are some additional factors that could contribute to a bubble burst in the AI startup space:
- A slowdown in economic growth: If the economy slows down, it could reduce investor appetite for risky investments like AI startups.
- Technological setbacks: If there are setbacks in AI development, it could erode investor confidence in the sector.
- Increased regulation: If the AI industry becomes more regulated, it could make it more difficult for companies to operate and grow.
My guess is the AI boom’s survivors will come in the second phase from those who have some sort of data speciality or user case an advantage. All generic AI API startups will fail to monetize enough and will be doomed.
What do you guys think?
#TecWinds #Chandana #DigitalNation
