Amazon + Anthropic join forces.
Written by: jh
Source: magke.com
Earlier today, Anthropic, a well-known unicorn in Silicon Valley, announced that Amazon would invest US$4 billion in it. The two parties announced a strategic cooperation to jointly promote the development of generative AI.
Just hearing the name Anthropic, some people may not be familiar with it. But you must have heard of their chatbot product Claude.
This chatbot, known as the strongest alternative to ChatGPT, is one of the few large-scale model products that can compete with ChatGPT. The most important thing is that it can be used for free (a paid version has recently been launched).
Anthropic itself is also closely related to OpenAI. Its founder is from OpenAI, and its financing documents directly state "benchmarking OpenAI."
It is such a company that was established less than 2 years ago and has been favored by many Silicon Valley giants including Google. Its valuation exceeded US$4 billion in the last round of financing.
Now that Amazon has invested US$4 billion, it is bound to create another AI giant.
$4 billion, where to spend it?
According to the press release, after Amazon completes its capital injection, Amazon employees and cloud computing customers will have early access to Anthropic's technology and integrate it into their own businesses.
Anthropic will choose Amazon Cloud Technology (AWS) as its main cloud provider and migrate most of its software to AWS data centers.
In addition, Anthropic will train and deploy future large AI models on Amazon's self-developed AWS Trainium and Inferentia chips - the huge cost of purchasing chips and training will be borne by Amazon.
In terms of details, Amazon will acquire a minority stake in Anthropic through an initial investment of US$1.25 billion, and any party has the right to trigger the remaining US$2.75 billion in capital injection.
However, Amazon claimed that its stake amounted to a minority stake and therefore they would not receive a seat on Anthropic's board of directors.
This may be a bit confusing. After spending so much money, you can’t get equity?
It is understood that Anthropic registered itself as a Public Benefit Corporation (PBC) from the beginning and established a long-term interest committee, which is composed of people who have no relationship with the company or its investors. They will include the board of directors. The company has the final say on matters including the composition of the company, in order to reduce the interference caused by commercialization.
This means that neither Google, which invested previously, nor Amazon, which invested US$4 billion this time, will have a direct impact on the company's future strategy. It also shows that Amazon will not be deeply bound to the Claude model.
Looking at it this way, the motivation for the cooperation between the two parties is actually very simple - to learn from each other's strengths and make progress together.
Hold on tight, AI stars can’t afford to spend money
There is no doubt that AI is the biggest trend this year. Among them, generative AI such as chatbots and AIGC shined in the technology circle in the first half of this year, and various product concepts continued to emerge.
On the surface, it looks very lively, but in fact it is difficult for ordinary people to have real access to generative AI tools, and there are no popular applications in enterprise business, and most of them stay on large models and product concepts.
The reason behind this, simply put, is that the threshold for generative AI is very high - it requires large amounts of computing power, high consumption costs, and a long period of time to make profits.
Not all companies can afford such a "gold-eating beast".
As far as Anthropic is concerned, although they have a luxurious team and excellent products, as an early start-up, funding has always been a big problem.
First of all, large language models require a lot of money and computing resources.
In order to create a large model comparable to ChatGPT, Anthropic invested a lot of money. It is reported that Anthropic is expected to spend US$1 billion on training and deployment this year, and will need a budget of US$5 billion to develop the next generation large model (Claude-Next).
According to Techcrunch statistics, the company's previous financing amount was approximately US$1.45 billion, which was far from enough for subsequent development costs.
Secondly, due to the characteristics of the Claude model (rejecting harmful prompt words and ensuring the safety of AI), Claude requires more cost to output results than other models.
From the consumer side, when the model performance is similar, Claude, which is priced better, obviously has no advantage. However, with its two major features of “free and powerful”, Claude has successfully received market feedback.
However, being free for a long time cannot afford to burn money. Anthropic has recently launched Claude Pro, a paid version of the Claude2 model, for about US$20 per month.
Judging from user feedback, this paid version has been opposed by the vast majority of people. After all, Claude’s selling point is “free”.
However, in the long run, Anthropic really cannot afford to burn money, and it is reasonable to launch a payment model.
With the arrival of Amazon's capital injection, Anthropic is finally expected to make up for the shortfall in subsequent development costs, and even complete training "for free."
In this way, Anthropic's free model should last for a while while ensuring the continuous upgrading of large models.
The AI war begins again
After talking about Anthropic, why did Amazon choose to invest heavily?
Simply put, they saw the potential of Anthropic's product, which is what Amazon AWS currently lacks the most.
In the second quarter of this year, Amazon delivered an excellent financial report. After completing large-scale layoffs, it finally got rid of huge losses and grew by 11% year-on-year.
However, behind the scenes, Amazon's proud AWS cloud service is still in a stage of slowing growth, which obviously does not satisfy executives and investors.
This is not the case for Amazon alone. The cloud services of giants such as Google and Microsoft have all been affected by the economic downturn. However, unlike Amazon, Google and Microsoft have continued to apply AI to their cloud business in the context of betting on AI, and have received a lot of rewards.
Here we have to mention the two AI giants DeepMind and OpenAI. They have achieved deep binding with Google and Microsoft respectively, and naturally they are still the cloud services of the giants.
On the other hand, Amazon, although it has the largest cloud service facilities, may be slightly behind in terms of generative AI layout.
Therefore, when Anthropic, the top AI company in Silicon Valley, encountered financial difficulties, Amazon naturally did not miss the opportunity to cooperate with it.
In fact, it is worth mentioning that in the terms of the transaction between Anthropic and Google, Anthropic regards Google Cloud as its "preferred cloud provider" and "jointly develops AI computing systems" with the two companies.
Now that Amazon has invested heavily in the game, it has naturally replaced Google Cloud's business.
On the other hand, as far as Anthropic's business model is concerned, they generate revenue by providing AI research and development services to enterprise customers and selling products and services that prioritize security. This is somewhat different from Claude's free model - yes, their long-term customers still prefer enterprise users.
Therefore, the benefits brought by Anthropic go far beyond themselves, as well as the customers of Anthropic services, which may bring benefits to Amazon cloud services.
In general, after the giants have invested in AI, a war on generative AI is quietly beginning.
In addition to Amazon, Google, and Microsoft, Tesla and Meta are strong competitors. And latecomers like Anthropic may become the biggest variable.
