Perplexity CEO warns on-device AI threatens $500 billion data center industry buildout

Is Perplexity CEO Right About the Threat to AI Data Centres? | AIM

Data Centre Bubble? Small AI Models May Paint a Different Future for GPUs Companies are shelling out billions of dollars in the race to buy GPUs and scale data centres, but that hyperscale hinges on a fundamental assumption: powerful AI models need centralised compute power to train complex algorithms. Now, some are challenging that very reality, hinting at the possibility that AI models may not need data centres at all. Aravind Srinivas, CEO and co-founder of Perplexity, in a recent podcast with Prakhar Gupta, argued that the biggest threat to data centres is local intelligence, where applications do not depend on compute hosted remotely. In this model, compute shifts closer to the user, reducing reliance on a centralised data centre-based infrastructure. Gavin Baker, CIO and managing partner at Atreides Capital, also echoed this view in a recent podcast. He imagined a future in which smartphones house more memory modules to accommodate pruned versions of frontier AI models, allowing users to access them without relying on cloud or high-end devices. Baker pointed to Apple's strategy, focused heavily on on-device, privacy-first AI rather than relying directly on powerful cloud-based models. That approach has improved privacy guarantees but limits massive data collection, contributing to Apple lagging in the broader AI ecosystem. Efficient and increasingly capable small language models strengthen the on-device case. Google continues to build large frontier systems such as Gemini 3 while also shipping the Gemma family of models. Smaller Gemma variants can run locally, and their performance has consistently improved across benchmarks and evaluations with each release. Paras Chopra, founder of AI lab Lossfunk, while testing a 270-million-parameter variant of Gemma, observed in a post on X that it was "absolutely wild how coherent and fast it is on my phone." Mobile applications such as PocketPal and Google AI Edge Gallery now allow users to download local models and experiment directly on smartphones. Google has also shipped on-device AI features across its Pixel lineup that do not rely on the cloud, prioritising speed and privacy. Beyond phones, developers have experimented with modified versions of powerful open-source models running locally on MacBooks with Apple silicon or on a single consumer GPU, achieving cloud-comparable results for specialised workloads. In the larger consumer device segment, NVIDIA is shipping compact Blackwell-based systems such as the DGX Spark workstation. PC manufacturers are also pushing AI PCs equipped with NPUs that can run AI workloads locally, albeit with limited features. The on-device thesis has its limits. Research institute Epoch AI stated in a recent report that using a "top-of-the-line gaming GPU like NVIDIA's RTX 5090 (under $2,500), anyone can locally run models matching the absolute frontier of LLM performance from just 6 to 12 months ago." "This relatively short and consistent lag means that the most advanced AI capabilities are becoming widely accessible for local development and experimentation in under a year," the report added. In a conversation with AIM, Sriram Subramanian, cloud computing analyst and founder of market research firm CloudDon, said he expects a mixed model, in which inference is split between the cloud and the device to improve performance. "The other angle is moving to smaller AI models where the requirements aren't much for the user." "GPUs will be the larger pie definitely," he declared, adding that powerful cloud-based compute will remain necessary for accuracy and high-demand workloads. But then there's the question of performance. If users want the most accurate and contextually relevant responses, they may continue to prefer cloud-based GPUs, which will remain more powerful than on-device systems, even as local AI proves increasingly capable. Minh Do, co-founder at Machine Cinema, a community of AI creatives, framed the trade-off in a post on X. "You wouldn't expect a poorly performing AI but a cheaper AI if the expensive one can accurately diagnose your grandmother or get all your math problems right." Moreover, AI models will continue to scale performance across tasks and domains. Rajesh C Subramaniam, founder and CEO of edge AI services company embedUR, told AIM that "what's changing [with on-device AI] is where inference makes the most sense." He explained that many edge AI workloads are situational and triggered by on-screen context or real-world interactions. These benefit from local processing due to latency, privacy, and cost. "In those cases, pushing inference to the device is simply the more efficient architectural choice." "At the same time, the cloud remains essential for tasks such as large-scale model training, fleet-level analytics, coordination across devices, and continuous improvement of models," he added. Moreover, hardware economics remain a constraint. DRAM prices are rising, which is expected to increase the cost of smartphones and laptops equipped with cutting-edge memory components to handle AI workloads. And eventually, memory pressure becomes acute for sensitive workloads such as facial recognition, payments, secure access, that involve data that should not leave the device. "In those scenarios, storing embeddings, reference data, and model parameters locally quickly becomes a challenge, especially on phones, laptops, or embedded platforms with strict memory budgets," Subramaniam explained. Security further shapes adoption. Carmen Li, CEO of Silicon Data, expressed that users may eventually be concerned over where their data is being processed -- on the phone or in data centres. She noted that trust depends on hardware-backed security like chip encryption technologies and their continued advancement. "Without that, you wouldn't feel that comfortable... majority of the users will be concerned," she added. Subramaniam also pointed to a talent bottleneck. "Modern edge hardware is capable, but extracting performance requires deep expertise in optimisation, quantisation, and hardware-aware deployment," he said. Small models may look feasible in theory. Deploying them reliably at scale remains a tricky proposition.

Perplexity CEO Says On-Device AI Can Disrupt the Data Centre Industry

AI lacks the biological drive to pose original questions, he said Perplexity CEO Aravind Srinivas highlighted a future where artificial intelligence (AI) shifts from centralised servers to local devices. During a podcast interview, he mentioned that the biggest threat to traditional data centres is if AI "can be packed locally on a chip that's running on the device" because that eliminates the need to run inference on it from data centres that power central servers. Srinivas also tackled other topics around the shift from chatbots to agents, the value of humans, and the future of the technology. Perplexity CEO Believes Powerful On-Device Intelligence Will Bring the Next AI Revolution During a podcast with YouTuber Prakhar Gupta, Srinivas opened up about his views on wide-ranging topics around AI. However, the highlight of the conversation was his views on data centres and locally run AI systems. "The biggest threat to a data centre is if the intelligence can be packed locally on a chip that's running on the device, and then there's no need to run inference on all of it on one centralised data centre," he said. The comment was made when he was asked by Gupta about a hardware upgrade that can bring a quantum leap to AI. Explaining his statement, the Perplexity CEO described a potential transition away from the prevailing model where most AI tasks are handled in large, specialised data centres. He suggested that if models were capable of on-device inference, the need for vast, centralised infrastructure could diminish. He said that such a shift would alter the economics of billions of dollars spent on data-centre buildouts worldwide and could lead to a more decentralised AI ecosystem. Beyond infrastructure, Srinivas also touched on the difference between biological and artificial intelligence, noting that human brains are far more energy-efficient than large data centres when measured per watt. He described human intelligence as driven by intrinsic curiosity, the capacity to question assumptions and investigate the familiar in new ways, something he said existing AI systems do not naturally exhibit. Srinivas also discussed how AI may influence the future of work and assistance. He suggested that ubiquitous, personalised AI could level the playing field between individuals and large institutions, much as smartphones have done by placing powerful tools in the hands of many people regardless of status. He noted that age is not a barrier to adopting AI tools but that a mindset of curiosity is critical to successful use.

Perplexity CEO Says On-Device AI Threatens Data Centers As Industry Faces '$10 Trillion Question' -- Apple, Qualcomm Positioned To Benefit - Alphabet (NASDAQ:GOOGL)

Aravind Srinivas, CEO of Perplexity AI, which is backed by Jeff Bezos and Nvidia Corp. (NASDAQ:NVDA), issued a contrarian warning about the future of artificial intelligence, saying on-device intelligence running on personal devices could disrupt the centralized data center model driving massive infrastructure investments. Localized AI Could Upend Data Center Industry "The biggest threat to a data center is if the intelligence can be packed locally on a chip that's running on the device and then there's no need to inference all of it on like one centralized data center," Srinivas said in a podcast interview with Prakhar Gupta released last week. Srinivas, who has previously worked at OpenAI, Google Brain, and DeepMind, said that AI running directly on personal devices could reduce the need for centralized data centers. "That really disrupts the whole data center industry like it doesn't make sense to spend all this money $500 billion, $5 trillion whatever on building all the centralized data centers across the world," he said, calling it a "$10 trillion question, hundred trillion dollar question." He also highlighted scenarios where AI running locally could learn from repeated tasks on individual devices, adapting over time and automating user activities. "It adapts to you and over time starts automating a lot of the things you do. That way you don't have to repeat it. That's your intelligence. You own it. It's your brain," Srinivas said. See Also: Apple Scales Back Vision Pro Production, Marketing After Sluggish Sales: Report Chip Companies and OEMs Positioned to Benefit Apple Inc. (NASDAQ:AAPL) has "a massive advantage" due to its M1 chips and power-efficient devices, according to Srinivas. Qualcomm Inc. (NASDAQ:QCOM) and original equipment manufacturers, including Apple, Samsung (OTC:SSNLF), Lenovo (OTC:LNVGF), and HP Inc. (NYSE:HPQ) could also benefit from distributing AI-enabled devices. However, technical barriers remain. Srinivas noted that no AI model has yet been released that can run efficiently on a local chip while completing tasks reliably. The Indian-born entrepreneur expects early adoption on MacBooks or iPads before reaching smartphones. Implications For Robotics And Labor He also discussed the potential for AI in the physical world, especially in robotics. Srinivas said AI could transform the labor market by automating many tasks now done by humans, echoing concerns raised by Geoffrey Hinton who is often called the 'Godfather of AI'. Industry Risks The U.S. economy is becoming increasingly reliant on AI, raising concerns about a potential AI bubble. If such a bubble bursts, centralized data centers could become a "single point of failure," with widespread economic repercussions. Srinivas's caution highlights a crucial question for the AI and tech industries: will centralized data centers continue to be the foundation of the digital economy as AI becomes more dispersed, or will intelligence on personal devices radically transform the sector? Read Next: AI Boom Creates Over 50 New Billionaires Amid Record $202 Billion In Funding Photo courtesy: Shutterstock Disclaimer: This content was partially produced with the help of AI tools and was reviewed and published by Benzinga editors. GOOGLAlphabet Inc$312.80-0.06%OverviewAAPLApple Inc$272.090.08%HPQHP Inc$22.330.24%LNVGFLenovo Group Ltd$1.4018.8%NVDANVIDIA Corp$186.46-0.02%QCOMQualcomm Inc$171.22-%SSNLFSamsung Electronics Co Ltd$42.48-%Market News and Data brought to you by Benzinga APIs

Perplexity CEO: On-Device AI Could Surpass Datacenter Reliance

Aravind Srinivas backs local and on-device AI as your own small brain In an era where tech giants are pouring trillions into massive data centers to fuel the AI boom, Perplexity AI CEO Aravind Srinivas is sounding a contrarian alarm: the real future of artificial intelligence lies not in the cloud, but on your device. Srinivas first gained attention for this view in a widely shared 2024 interview, where he declared on-device AI the "biggest threat" to centralized data centers. "If the intelligence can be packed locally on a chip that's running on the device," he argued, "there's no need to inference all of it from one centralized data center. It becomes more decentralized." He envisioned models adapting to users through "test-time training," observing repeated tasks, retrieving local data on-the-fly, and automating workflows, all while keeping everything private. "That way, it's your brain," he said, warning that unchecked data center buildouts could prove wasteful. Also read: Should AI get legal rights? It's dangerous for humans, warns expert Throughout 2025, Srinivas reiterated and expanded on this prophecy in interviews, podcasts, and public appearances. He emphasized privacy as a foundational advantage: data remains on the user's device, eliminating the need to send sensitive information to remote servers. "All your data lives on your client. We don't need to take any of it," he has stressed, pointing out vulnerabilities in cloud-dependent agents that require ongoing authentication. Advancements in efficient models and specialized chips from companies like Apple, Qualcomm, and Arm have brought this vision closer to reality. Srinivas draws historical parallels to the transition from mainframes to personal computers, forecasting a similar shift where AI companies compete by shipping highly optimized, device-native models rather than relying on ever-larger cloud infrastructure. Also read: Meta's big AI play: What the Manus acquisition means for automation at scale At Perplexity, this philosophy is driving product development, including the Comet AI browser and upcoming desktop experiences. Srinivas has outlined ambitions for local models that deliver fast, battery-efficient performance for tasks like browser control, email management, and multi-tab research - without latency or privacy trade-offs. "If we can make models small enough, fast enough, and power-efficient enough to run locally - without draining your battery or compromising intelligence - that's the true magic," he explained in 2025 discussions. This approach challenges the dominant paradigm of hyperscalers like Microsoft, Google, and Meta, who continue massive GPU investments for frontier-scale training. Critics note that cutting-edge capabilities still demand cloud resources, with on-device models often smaller and limited for complex queries. Srinivas counters that for personalization, real-time interaction, and privacy-critical use cases, the bulk of daily AI value, local execution wins out. As 2026 dawns, the debate intensifies: will we see a hybrid ecosystem, or a genuine pivot to edge AI dominance? Srinivas's bold stance suggests the latter could reshape the industry sooner than expected.