‘Almost unlimited’: Execs says AI demand remains strong even as enterprises move to ‘valuemaxxing’

The semiconductor market has experienced a remarkable surge over the past year, fueled by optimistic investments in AI infrastructure. However, recent fluctuations in chip stock prices have raised concerns about the sustainability of this demand. Despite these uncertainties, several AI leaders have confidently asserted that interest in AI technology remains robust, albeit with a more cautious approach regarding expenses. Pat Gelsinger, former Intel CEO and current general partner at Playground Global, expressed his unwavering belief in the boundless potential of AI demand. "I consider AI demand to be almost unlimited," he stated in an interview with CNBC, identifying energy availability as the primary constraint. He emphasized the immense economic value that enhanced intelligence brings across various industries, dubbing it nearly infinite. Market volatility has been exacerbated by several factors, including Meta's announcement to sell its surplus AI computing resources, which stirred concerns about potential overcapacity in the market. While Meta's stock initially benefited from this news, it sparked broader questions regarding the industry's overall capacity. Similarly, Elon Musk's xAI has also begun renting out its excess computing power. In contrast, Samsung, a leading memory chip manufacturer, recently forecasted a significant profit increase, yet its stock declined, reflecting skepticism about the extent of its recent 360% share price increase over the past year. Despite these market dynamics, the appetite for computing infrastructure remains unyielding. Marc Boroditsky, chief revenue officer at Nebius, noted, "The demand we're witnessing is extraordinary, surpassing our current capabilities for fulfillment." Nebius is actively developing data centers equipped with Nvidia's GPUs to meet this rising demand. Andrew Feldman, CEO of Cerebras Systems, pointed out that the cases of Meta and xAI selling excess capacity are not indicative of the overall industry trend. "The demand for computing resources significantly exceeds the available supply, and we are experiencing a shortage of data centers," he explained. Sungyun Park, CEO of Rebellions, a South Korean chip startup, echoed this sentiment, asserting that AI infrastructure demand continues to thrive. "The current momentum in AI infrastructure is substantial," he remarked, dismissing notions that hyperscalers are overextending their investments. Lumentum, a company specializing in photonics and optical products, reported a complete sellout of its offerings for the next five years. CEO Michael Hurlston stated, "We are ramping up our production capacity to meet the anticipated demand over the next five years." The conversation surrounding AI is also shifting towards the financial implications for enterprises. Some companies have engaged in a practice dubbed 'tokenmaxxing,' encouraging widespread AI usage regardless of outcomes. However, many are now recalibrating their focus towards maximizing returns on investment from AI technologies, particularly as the costs of advanced models remain high compared to open-source alternatives. Boroditsky emphasized that the push towards rational spending is vital, stating, "Organizations should prioritize value generation from AI investments. The demand for AI solutions will continue as companies adapt to these economic realities." As enterprises refine their approaches to AI, the landscape is likely to evolve, with specific models becoming more suitable for particular tasks. Feldman concluded, "As we advance in our understanding and deployment of AI, we will see a more efficient allocation of resources depending on the complexity of the tasks at hand."