How much have the biggest companies invested in AI and what has it really cost them?
Training AI alone can cost an organization over a hundred million dollars, as it did with GPT-4. Many of the costs that companies face when investing in machine learning are not taken into account. However, they can be so high that they create a huge gap between the costs and the benefits of AI. Let's look at how much corporations are actually spending on AI, whether these costs are justified, and why AI is so expensive that it could eventually create a monopoly in the market.
What are AI investments made up of and when did they pay off?
The first acquisition of an AI startup was in 1998, when Microsoft bought Firefly Network for just $40 million. The company's main product was used to anonymously collect user preferences, recommend content, and send advertising. Over the next 20 years, various corporations spent nearly $8.6 billion in such acquisitions. But acquisitions and mergers represent only a small portion of AI investment.
Back in the 2010s, when there was less hype around AI and companies were investing in AI startups wisely, the biggest investments and deals often paid off and led to the creation of entire divisions in today's IT giants and AI industry leaders.
For Google, its biggest acquisition in terms of AI startup investments during that time was Nest Labs purchase for $3.2 billion. The company created equipment for smart homes and one of their most successful developments was a smart thermostat. It collected data based on user experience and began to independently program the temperature preferred by the home owners, also automatically adjusting to the change of time of day and year. Gradually, the company formed the basis of the Google Nest brand – a division of consumer electronics.
Amazon in 2012 implemented the product of the startup Kiva Systems. These are robots that move around the warehouse, handing over goods to employees. The startup was renamed Amazon Robotics after being acquired for $775 million. The technology has had a huge impact, allowing it to manage huge volumes of orders with unprecedented precision and without interruption.
Google in 2014 acquired Another significant startup is Deep Mind for $400 million. Initially, this company developed artificial intelligence that beats humans in chess and other games. In 2016, the company taught a neural network to play 3D games. It was their product that was able to beat champions in Go – this game, unlike chess, does not have classic combinations and repeating games. Later, the company began developing AI for image generation and AI for medical purposes. The company still exists in the Google DeepMind format.
At the time, these were unprecedentedly large deals. As AI began to gain popularity in the information field, corporations greatly increased their investments. Microsoft alone invested more than $ 10 billion. Changes are also noticeable in the field of startup funding. In 2023, Anthropic alone attracted $6 billion. The approach to investing in AI has changed dramatically, and companies are betting really high on the technology. The question is, are these investments paying off?
Who benefits from investing in AI?
Grand View Research estimated the global AI market in 2023 by $196.63 billion. They predict that this area will grow at a rate of 37.3% from 2023 to 2030.
McKinsey estimates that artificial intelligence can increase corporate profits by $4.4 trillion per year. They also expect a positive impact of AI on labor productivity – by 0.1-0.6% annually until 2040. But this will only be useful for the economy if the employees affected by the technology find other activities.
Microsoft has already seen significant returns on its AI investments, including OpenAI. Leaders in 2023 statedwhich expects $10 billion in annual profits from AI alone. Shares responded by jumping sharply, reaching their highest since 2021.
Nvidia is a big beneficiary of the growth of the AI market and the investments of companies in this area, as this corporation produces chips for artificial intelligence. Thus, according to the report for 9 months of 2023, out of $18.12 billion in revenue, $14.51 billion was earned by the data center division.
Nvidia founder and CEO Jensen Huang himself has cited NVIDIA's GPUs, networking technologies, AI foundry series, and Nvidia AI enterprise software as the company's growth engines. The company makes a lot of money from AI-tailored products, such as GPUs tailored for AI generator workloads.
How much profit is needed to pay for processors
Given the boom in artificial intelligence and the attention it receives, the issue of investing in AI hardware is very acute. Companies are investing huge amounts of money just to keep up with the technology, even if the chances of “recouping” these investments are not that high.
Sequoia Capital analyzed the spending of cloud service providers and other companies. According to calculations, They could spend $300 billion on AI hardware in 2024This is 1.5 times higher than global investment in 2023.
The company will spend $150 billion on purchasing graphics processors from Nvidia alone. Data center costs (electricity, buildings, and backup generators) will require investments roughly equal to the cost of the processors;
Software returns average 50%, so to recoup a $300 billion investment, companies would need $600 billion in revenue.
Such superprofits may take decades to achieve. What may happen in the coming years is that revenues will increase enough to keep corporations growing, and their stock prices will continue to grow.
Similar analytics at The Economist. They calculated that IT giants like Alphabet, Amazon, Apple, Meta (banned in Russia, like Facebook in one of the diagrams above) and Microsoft are planning about 400 billion in capital expenditures on AI, including equipment, as well as research and development. At the same time, the combined market capitalization of these five companies has grown by 2 trillion dollars over the past year. Such indirect profits and revenues from AI are of great importance. However, the already mentioned 10 billion dollars of Microsoft's income from AI does not seem so significant if you know the equipment costs. And yet this is one of the most successful examples and optimistic expectations.
What revenues do companies receive from AI now?
So far even Top 10 AI Companies by Revenue do not receive such high incomes, and if this happens, it is due to the specifics of the business.
OpenAI is estimated to receive approximately $1.3 billion per year;
Language model developer Anthropic is estimated to have achieved annual revenues of $850 million;
Microsoft through collaboration with OpenAI, Bing Chat products and the Azure cloud, it is capable of generating 10 billion in revenue per year;
Nvidia. As already mentioned, it makes money on AI development. The data center division receives revenue at the level of $15 billion;
Hugging Face. The platform could reach revenues of $30-50 million per year;
Stability AIcreator of the Stable Diffusion image generator. The company earned $1.2 million and was valued at $1 billion;
Perplexity AIan AI-powered search engine valued at half a billion and bringing in $5-10 million annually;
IBM. Launched virtual assistant Watsonx, a data processing platform. They thinkthat the company's AI business is approaching a billion dollars in annual revenue;
Google. Google's research assistant Bard and the multimodal LLM family Gemini were launched. Exact AI revenues are unknown;
Salesforce. Following announcements of intentions to invest in AI as part of a CRM subscription.
Of these companies, only Microsoft, IBM, and Google are among the top corporations in terms of AI spending. And even then, their revenues have not justified these expenses.
Tom's Hardware estimates that companies will make about $100 billion in profits from AI in 2024. Giants like Google, Microsoft, Apple will make about $10 billion each, if optimistic forecasts are knocked down, and companies like Oracle, ByteDance, Alibaba, Tencent, X, and Tesla will make $5 billion each. However, this is a positive estimate; The Economist suggests a more significant gap between costs and revenues.
Is the AI bubble dangerous and are we headed for another 'dot-com crash'?
If this gap between spending and revenue is not closed, a bubble could emerge. But it’s not all doom and gloom. Most of the companies investing heavily in AI hardware are funding it with their own cash, not debt. Some, like Microsoft, are so profitable that their investors don’t worry about such spending.
What investors fear more is that their company will be at the back of the race for a century-defining technology. Microsoft, for example, had a large cash position ($111 billion), a high profit margin (34%), and relatively little debt in 2023. Such companies can take on such expensive risks as AI.
Amazon, for example, is in a slightly different situation. It does not have such a high net margin, as well as significant long-term debt. Thus, in 2023, the corporation had $ 87 billion in cash and $ 136 billion in long-term debt. The company's profitability and margin are also not that high. As a result, Amazon needs more profit to justify investments in AI.
At the same time, all these organizations declare their expenses on AI, but do not allocate profits from it as a separate item, even with rough estimates. It will take too much time before the income becomes sufficiently pronounced and significant for this.
It may take 10 or 20 years before a “killer” AI product emerges that can be actively marketed or otherwise monetized. This has happened with other disruptive innovations. The first personal computer was introduced to the public in 1971, but it was eight years before the first spreadsheet hit the market, which for many became the cause buy this PC. Or another example. The iPod appeared in 2001, but until the iTunes Store was launched in 2003, it was not particularly popular.
So, a bubble is forming in the AI market. But as long as it is not being inflated with borrowed money, the consequences should not be as large-scale as during the Dot-com Crash. It is believed that companies take on primarily “technical debt”, that is, expected costs that will have to be incurred in the future. This is infrastructure modernization, fixing vulnerabilities, etc. In addition, we can talk about “debt on expected income”. High expectations are formed on the stock market from companies investing in AI. But as already said, they may not be able to “recoup” these investments in the short or even medium term. The subsequent disappointment, the decline in the “hype”, will lead to a collapse of shares and subsequent shock in the market.
How the High Cost of AI Leads to Monopolization
Why do companies continue to invest in AI despite everything? Whether these estimates are overstated or not, it is believed that by the end of the decade, AI will have a huge impact both on the market and in everyday life. Its use in corporate processes can significantly increase the efficiency of an organization, as it once happened with Amazon after the introduction of AI robots in the 2010s. Such solutions can provide such a large competitive advantage that a company can become a virtual monopolist in the market.
PWC provide datathat by 2030, 45% of economic growth can be achieved through product improvements. Moreover, this is attributed to the fact that AI helps expand the range of products and improve the personalization of services. Companies implementing AI will be able to offer customers much more and better.
Another problem is that, in the end, the AI market is actually formed by only a few companies. This opens up additional opportunities for monopolization of the AI sphere. It will also increase the gap between those who control AI and those who are affected by it. The high cost of creating AI, including in terms of equipment costs, makes the barrier to entry into the market very high, and ultimately, if AI becomes the defining technology of the future, then IT giants will control it. They will also determine the rules of the game. You don’t have to go far for examples. Let’s recall the history of Standard Oil and how it set its own prices and crushed competitors like flies. At the beginning of the 20th century, it had to be forcibly split into several smaller enterprises. Perhaps something similar awaits a hypothetical Microsoft in the future. The question is, in what future? In the near or distant future?
Results
The AI industry requires companies to spend a lot of money in many ways, and one of the main ones is to provide high-tech equipment. So far, companies are spending hundreds of billions of dollars on AI development, but it is not the IT giants who benefit from this, but the equipment suppliers, mainly the same graphics processors. High costs and lack of return are creating a bubble in the market that may burst in the future.
There are other consequences, too. Since AI requires large investments, but pays off far from immediately, only a limited number of companies with inflated budgets can create competitive products and innovations. This will potentially lead to a monopolization of the AI market, with all the negative consequences. Say hello to Standard Oil founder John D. Rockefeller and his fortune of 1.5% of US GDP. Only in the 21st century did humanity get new “oil,” but wells have only just begun to extract it.
Free search, monitoring and registration of trademarks and other intellectual property objects.
More content about intellectual property in our Telegram channel
