SAN FRANCISCO: A New Era of AI Resource Sharing
The foremost artificial intelligence enterprises are allocating colossal sums—totaling in the hundreds of billions of dollars—toward the construction of data centers essential for AI development.
Technology behemoths such as Amazon and Google, alongside financially robust startups including Anthropic and OpenAI, constitute a select cadre of organizations equipped with the capital and connections necessary to access this computing prowess.
This scenario effectively marginalizes a multitude of other entities.
Anjney Midha, a seasoned tech entrepreneur previously aligned with the venture capital firm Andreessen Horowitz, aspires to alter this status quo through a novel startup.
His fledgling company, Amp, is endeavoring to acquire surplus computing capabilities from data center operators across the United States and beyond, thereby redistributing this invaluable resource to those in need.
Headquartered in Menlo Park, California, Amp seeks to establish a global reservoir of specialized microchips accessible by startups, educational institutions, and various organizations that traditionally lack the requisite computational resources to train cutting-edge AI models.
“Certain companies simply cannot secure the computing capabilities they require,” Midha remarked. “The most affluent and influential companies are monopolizing the infrastructure for their exclusive benefit.”
Amp has successfully garnered over US$1.3 billion (RM5.16 billion) from a range of investors, including Andreessen Horowitz, the startup incubator Y Combinator, and multiple cloud computing firms.
Several noteworthy startups have also pledged to utilize and share from Amp’s pool of computing prowess. These include Periodic Labs, an AI enterprise concentrated on scientific exploration, and Eleven Labs, which specializes in developing AI systems for voice synthesis.
Amp’s initiative is part of a broader movement aimed at consolidating AI infrastructure. Chipmaker Nvidia and French startup Mistral recently announced a collaboration to pool computing resources for the development of AI systems tailored for European companies, aiming to diminish reliance on U.S. tech giants.
Midha drew an analogy between his startup and the establishment of an electrical grid. Just as electricity can be disseminated among homes and businesses, a collective pool of computing capabilities can be shared among startups and academic institutions.
Investors supply capital that Amp utilizes to procure computing power from data center operators. Subsequently, AI startups affiliate with the coalition to leverage this computing muscle for model development. In return, these startups contribute supplementary funds or alternative resources.
Some startups might share essential digital data required for AI model training, while others could reciprocate by providing their software code or collaboratively training AI models.
The paramount advantage of such collaboration lies in collective bargaining, stated Liam Fedus, CEO of Periodic Labs. Individually, his startup would face significant challenges securing the necessary computing power.
However, if Amp can negotiate with data center operators on behalf of multiple startups, it can wield greater influence.
“Pooling our demand enables us to engage in substantially more serious discussions about acquiring computing resources,” Fedus remarked.
Source link: Thestar.com.my.