How DeepSeek Changed AI Forever and What It Means for Compute Power
In Brief
DeepSeek’s $5.6 million model training revolutionized the AI market, challenging the belief that building powerful AI costs billions. However, this success may signal a rise in compute power demand.
The recent success of DeepSeek, a Chinese AI startup that trained a model equal to GPT-4 for only $5.6 million, shook up the AI market. This breakthrough called into question the premise that building powerful AI costs billions of dollars. While some saw this as a hint that AI compute costs may be reduced, the opposite is most likely true. DeepSeek’s success indicates an oncoming rise in compute power demand, which has the potential to disrupt the whole AI ecosystem.
Why Cheaper Computers Drive Demand?
To comprehend the future of AI compute, consider Jevons’ Paradox. In the nineteenth century, economist William Jevons recognized that advancements in steam engine efficiency resulted in increased coal consumption. Lower costs allowed for greater utilization, driving overall demand higher.
A similar pattern applies to AI compute. As the cost of training sophisticated models decreases, more businesses and developers will enter the area:
- Startups that previously couldn’t afford AI may now explore ambitious ideas.
- Businesses that are afraid to automate their operations will adopt AI technologies.
- Industries without a clear AI use case will begin to investigate applications.
When access expands, demand for computational resources is expected to increase dramatically. Morgan Stanley predicts that AI compute demand will double every six months, implying that by 2030, training an artificial general intelligence (AGI) system might consume as much energy as a small nation.
A Market Dominated by Big Tech
The AI compute industry is now dominated by three companies: Amazon Web Services, Google Cloud Platform, and Microsoft Azure. While these corporations supply critical infrastructure, their dominance presents obstacles for smaller players and innovation in general.
High-performance GPUs, such as Nvidia’s H100 and H200, are necessary for training modern models. However, these resources are frequently set aside for Big Tech’s internal initiatives or its major clients. Developers and entrepreneurs typically encounter lengthy wait periods or are refused access entirely.
Even when access is accessible, the cost may be excessive. AWS charges about $14 per hour per H100 GPU, whereas GCP charges roughly $6-7 per hour. These expenses impose a barrier on smaller enterprises and researchers, reducing their capacity to compete with well-funded corporations.
Building an AI-dependent company on Big Tech infrastructure entails considerable risk. Companies that spend extensively on model training may face access restrictions or additional prices. Deep collaborations between cloud providers and top AI developers, such as Microsoft’s collaboration with OpenAI, heighten worries about favoritism and access restrictions.
The concentration of computational capacity in the hands of a few businesses emphasizes the necessity for independent suppliers. Exabits is emerging as one such option, with the goal of making AI infrastructure more accessible to everyone.
Exabits distinguishes itself by providing high-performance compute systems optimized for AI applications. Unlike some decentralized compute initiatives that struggle with stability, Exabits offers GPU clusters capable of performing large-scale AI operations.
Recognizing that AI compute demand will continue to rise, Exabits is boosting its capacity across different locations. This technique assures continual availability, allowing firms to grow seamlessly.
The comany promises to make compute power more cheap. While specific pricing vary by usage and geography, the firm positions itself as a less expensive option to AWS and GCP. Lower costs may allow startups, SMEs, and researchers to pursue AI development without financial constraints.
Securing access to high-end GPUs remains a top priority. Exabits provides a variety of hardware, including the Nvidia H100 and H200, as well as additional models like the 4090, A100, Ade6000, and MI50. The availability of approximately 60,000 GPUs, including 3,000 H100 and 4,000 H200 units, provides customers with the resources they need to train complex models.
Implications of an Open AI Compute Market
As AI capabilities advance to AGI, computational power will play a critical role in molding the future. If Big Tech retains its dominant position, it may effectively limit access to AGI development and deployment. This raises questions about competitiveness, innovation, and the fair sharing of AI gains.
In contrast, a diverse computational environment may promote a more open and inclusive future. Companies like Exabits provide an alternate framework in which startups, researchers, and companies all have equitable access to the infrastructure needed for AI innovation.
Preparing for the AI compute Surge
DeepSeek’s accomplishment does not imply that AI research will become less resource-costly. It suggests that AI is becoming more accessible, resulting in increased demand for processing resources. This growth pattern is compatible with Jevons’ Paradox, which claims that efficiency gains will lead to increased adoption and consumption.
Big Tech’s existing grip over AI infrastructure creates hurdles for startups and the industry as a whole. Limited access, high expenses, and dependence hazards underscore the importance of alternative suppliers. Exabits stands itself as a viable option with enterprise-level performance, scalability, low price, and access to cutting-edge GPUs. Companies like Exabits might play an important role in crafting a future in which AI development is not limited to a select few but is available to innovators all around the world.
Disclaimer
In line with the Trust Project guidelines, please note that the information provided on this page is not intended to be and should not be interpreted as legal, tax, investment, financial, or any other form of advice. It is important to only invest what you can afford to lose and to seek independent financial advice if you have any doubts. For further information, we suggest referring to the terms and conditions as well as the help and support pages provided by the issuer or advertiser. MetaversePost is committed to accurate, unbiased reporting, but market conditions are subject to change without notice.
About The Author
Victoria is a writer on a variety of technology topics including Web3.0, AI and cryptocurrencies. Her extensive experience allows her to write insightful articles for the wider audience.
More articles
Victoria is a writer on a variety of technology topics including Web3.0, AI and cryptocurrencies. Her extensive experience allows her to write insightful articles for the wider audience.
