The Cost of Intelligence is Approaching Zero
Imagine a world where computational power and sophisticated AI models become virtually free and available to everyone. This is not hypothetical; it's rapidly becoming our reality. Let's explore the three powerful forces driving the cost of intelligence to zero.
Reason 1: Massive Cloud Infrastructure Growth
The global cloud computing landscape is undergoing unprecedented expansion, driven significantly by advancements in AI. In 2023, the market was valued at approximately $626.4 billion and is projected to reach $1.3 trillion by 2028. This growth is propelled by substantial investments from leading technology companies aiming to enhance their cloud and AI infrastructures. For instance, Amazon plans to allocate $100 billion in capital expenditures in 2025, primarily focusing on AWS and AI infrastructure. Microsoft has announced an $80 billion investment in AI and data centers for its fiscal year 2025. Similarly, Alibaba has pledged $53 billion to expand its AI and cloud infrastructure.
In addition to these investments, NVIDIA, a key player in AI hardware, has committed to a millionfold increase in computing power over the next decade, primarily directed toward cloud resources. Complementing these efforts, the Stargate Project—a joint venture involving OpenAI, SoftBank, Oracle, and MGX—plans to invest up to $500 billion in AI infrastructure in the United States by 2029.?
Reason 2: Dramatic Drop in Model Training Costs
The second driver behind dropping AI costs is the drastic reduction in expenses associated with training and running sophisticated AI models. Take DeepSeek R1, for instance—it shocked the AI world by achieving a training cost of just $5.5 million, vastly cheaper than OpenAI’s estimated $100 million cost.
Ironically, regulatory restrictions intended to slow China’s AI advancements prompted this innovation. DeepSeek bypassed conventional, fully dense transformer models, opting for a Mixture of Experts (MoE) architecture, dramatically reducing the computational requirements and costs.
The rapid evolution continued impressively with Stanford University's creation of the Stanford S1 model, released only 11 days after DeepSeek R1. Stanford researchers astonishingly trained their model in just 26 minutes on 16 H100 GPUs at an astounding cost of merely $50. Stanford's method introduced a novel "budget forcing" technique.
And the innovation hasn't stopped there. Inception Labs introduced a Diffusion-Based LLM this month, claiming to be 10 times faster and 10 times cheaper than previous technologies, promising continued acceleration of this cost decline.
Reason 3: Open-Source Revolution
The final crucial factor propelling intelligence costs toward zero is the rapid growth of the open-source movement. Both DeepSeek R1 and Stanford S1 not only dramatically reduced training and execution costs but also made their revolutionary models freely available as open-source.?
This open-source shift is transformative. In 2024, 95% of organizations reported increased investments in open-source AI initiatives, marking a clear industry-wide shift. The open-source AI market is expected to grow to $50 billion by 2026.?
Industry leaders like Elon Musk and Sam Altman have hinted at releasing older versions of their proprietary AI models into open source as newer versions become commercially available. This move democratizes access to advanced AI technologies, accelerating adoption and integration across various sectors.
Conclusion: Implications for the Future
We now stand at the intersection of three significant trends: massive growth in cloud computing infrastructure, sharply decreasing model training costs, and the broad availability of powerful open-source AI models. Collectively, these trends will drive the price of intelligence steadily toward zero.?
And with that, here is the latest news in AI from last week.
Notable AI news from Last Week
Judge rejects Musk’s effort to prevent OpenAI from going for-profit entity:? A federal judge in California has denied Elon Musk's request for a preliminary injunction to block OpenAI's transition to a for-profit model, stating that Musk did not meet the required burden of proof for such extraordinary relief. However, the court is prepared to expedite a trial to address Musk's remaining claims against OpenAI and its CEO, Sam Altman, given the public interest at stake.?? ?
OpenAI Plots Charging $20,000 a Month For PhD-Level Agents:? OpenAI reportedly plans to introduce specialized AI agents, with monthly pricing tiers ranging from $2,000 to $20,000. These agents are designed to perform tasks such as sorting sales leads, software development, and supporting PhD-level research. ?
Eric Schmidt argues against a ‘Manhattan Project for AGI’:? Former Google CEO Eric Schmidt, Scale AI CEO Alexandr Wang, and AI safety expert Dan Hendrycks argue against a U.S.-led "Manhattan Project" for superintelligent AI, warning it could provoke retaliation from China and destabilize global relations. Their paper, Superintelligence Strategy, advocates a defensive approach, proposing measures like cyber deterrence and restricting adversaries’ AI access rather than racing to dominate AGI.? ?
The US Department of Labor is investigating scale AI:? The U.S. Department of Labor is examining Scale AI for potential violations of labor laws, including misclassification of workers and unpaid wages. The probe, active since August 2024, does not imply wrongdoing but follows legal challenges from former workers alleging underpayment. Scale AI, valued at $13.8 billion, denies any violations, emphasizing its role in providing flexible AI-related work.? ?
The Humane Ai Pin Has Already Been Brought Back to Life:? Humane's AI Pin, a widely criticized AI-enhanced wearable, was officially shut down on February 28, 2025, rendering the $700 device useless. However, a group of dedicated users and hackers refused to let it die, working together on Discord to find ways to bypass Humane’s authentication and restore functionality. With no official support from Humane, they reverse-engineered the hardware, created custom connectors, and ultimately gained access using an internal certificate leaked by a former employee.??
Health experts warn of leaning too heavily on AI for social connections:? AI companions are growing in popularity as online friends or romantic partners, raising concerns about their impact on real-world social connections. Social health expert Kasley Killam acknowledges AI's potential to help practice social interactions but warns against replacing human relationships. Speaking at SXSW, she stressed that real-world social skills should be nurtured in person, not just with AI.? ?
AI Thinks It Cracked Kryptos. The Artist Behind It Says No Chance:? For 35 years, cryptographers have tried to crack Kryptos, a sculpture at CIA headquarters, but its final section, K4, remains unsolved. AI has fueled a wave of misguided solutions, with people convinced they've cracked it using chatbots. Artist Jim Sanborn, the only person who knows the answer, is frustrated by AI-generated claims that lack real cryptographic effort. Despite dropping small hints over the years, K4 remains elusive. Sanborn, now nearly 80, refuses to reveal the solution in his lifetime, hoping Kryptos remains an enduring mystery. ?
The US Army Is Using ‘CamoGPT’ to Purge DEI From Training Materials:? The US Army is utilizing an AI tool called 'CamoGPT,' initially developed to enhance productivity and operational readiness, to systematically remove diversity, equity, inclusion, and accessibility (DEI) content from its training materials, aligning them with directives issued by President Trump. ?
Honor's New AI Agent Can Read and Understand Your Screen:? Honor has unveiled its new AI-driven feature, the Honor UI Agent, at the Mobile World Congress 2025 in Barcelona. This innovative tool allows users to interact with their devices more seamlessly by understanding and manipulating on-screen graphical user interfaces (GUIs). For example, during a demonstration, the agent autonomously booked a restaurant table through the OpenTable app by interpreting the screen's content and performing the necessary actions without external API calls.
McDonald’s Gives Its Restaurants an AI Makeover: McDonald's announced plans to upgrade its 43,000 restaurants with AI-powered drive-throughs, internet-enabled kitchen equipment, and management tools to improve customer and employee experiences.
China’s Autonomous Agent, Manus, Changes Everything: China's recent unveiling of Manus, an autonomous AI agent developed by the startup Monica, marks a significant advancement in AI. Launched on March 6, 2025, Manus is designed to independently analyze, plan, and execute complex tasks without human intervention, positioning it as one of the world's first fully autonomous AI agents. Its capabilities span various domains, including analyzing financial transactions, screening job applicants, and sourcing suppliers, delivering faster and more efficient decision-making than humans.
Recent Investment Activity
Tracking 31 AI companies last week that raised $4.12B, here are the highlights:
AIceberg Based in New York and five years old, AIceberg offers a platform to monitor and control AI interactions, helping prevent risks such as data leaks. The company raised $10?million in a seed round led by SYN Ventures and Sprout & Oak.
Aescape An eight-year-old New York startup, Aescap, develops AI-powered robotic massage tables for wellness and fitness facilities. It secured $83?million in a funding round led by Valor Equity Partners.?
Alta Alta is a two-year-old Tel Aviv startup providing AI tools that automate sales and revenue tasks for small and medium-sized businesses. The company raised $7?million in a seed round co-led by Entrée Capital and Target Global.
Anthropic Anthropic, an AI startup founded in 2021 by ex-OpenAI employees, raised $3.5 billion in a funding round led by Lightspeed Venture Partners, boosting its valuation to $61.5 billion.
Anysphere A three-year-old San Francisco startup, Anysphere offers an AI-powered coding assistant that suggests code and aids in debugging. It is reportedly preparing to raise a substantial amount at a valuation of around $10?billion.
Archipelo Based in San Francisco and five years old, Archipelo tracks actions by developers and AI coding tools to enhance code security and compliance. It raised an $8?million seed round and has a total of $12?million raised.?
Ataraxis AI Ataraxis AI, a New York-based startup, uses artificial intelligence to predict cancer outcomes by analyzing patient data. It raised $20.4?million in a Series A funding round led by AIX Ventures.
Auxia A three-year-old Palo Alto startup, Auxia provides an AI-driven marketing platform that creates personalized customer journeys for large enterprises. It has raised $23.5?million in seed and Series A rounds.
Breakout Breakout is a one-year-old San Francisco startup offering an AI-powered sales representative to manage early customer interactions. The company raised $3.25?million in a seed round led by Village Global.
Copley This one-year-old Boston startup uses AI to help brands create, test, and distribute marketing content across channels. Copley secured $4.8?million in a funding round co-led by Asymmetric Capital Partners.
dScribe AI dScribe AI, a one-year-old startup from Kansas City, analyzes various types of videos to track trends and improve internal knowledge management. It raised $1.2?million in a pre-seed round with support from several early-stage investors.
Exia Labs Based in Bellevue, WA, Exia Labs develops AI as a co-pilot for military planning by automating decision-making and simulating strategies. The startup raised $2.5?million in a pre-seed round led by Andreessen Horowitz.?
FairPlay FairPlay is a four-year-old Los Angeles startup that helps organizations detect and correct biases in their AI systems. It raised $10?million in a funding round co-led by Infinity Ventures and JPMorganChase.
Firsthand A two-year-old New York startup Firsthand, builds an AI-powered platform that enables brands and publishers to engage consumers through custom AI agents. It raised $26?million in a Series A round led by Radical Ventures.?
Heidi Health Heidi Health is a six-year-old Melbourne startup that offers AI-powered software to transcribe doctor-patient conversations into clinical notes. The company raised $16.6?million in a Series A round led by Headline.
HeyMilo Based in New York and two years old, HeyMilo uses AI to automate the early stages of candidate interviews with interactive voice assessments. It secured $2.2?million in a funding round led by Canaan Partners.
InsureVision A two-year-old UK startup, InsureVision leverages AI to analyze video from vehicle-mounted cameras to predict and prevent car accidents. It raised $2.7?million in a seed round with backing from several investors.?
Installer Installer is a three-year-old Oslo startup that employs AI to streamline the installation and maintenance of renewable technologies across Europe. The company raised $4?million in a seed round led by Brighteye Ventures.
Jurimesh A two?year?old Ghent startup that automates legal due diligence by using AI to detect potential issues and generate detailed reports. It raised a $1.7M pre?seed round led by Syndicate One.?
Kleio Based in Paris, Kleio uses AI to interact with potential buyers, answer questions, and guide purchasing decisions. The startup raised $3.15?million in a seed round co-led by Serena Capital and Daphni.?
Knostic A two-year-old Tel Aviv startup, Knostic provides security tools to control access to information in AI systems. It has raised $14?million, including an $11?million round led by Bright Pixel Capital.
LlamaIndex LlamaIndex is a three-year-old San Francisco startup that enables companies to connect diverse data sources to large language models. It raised $19?million in a Series A round, totaling $27.5?million.
MaxIQ A three-year-old startup from San Ramon, CA, MaxIQ uses AI to help B2B SaaS companies track and predict customer behavior. It raised $7.8?million in a seed round led by Dell Technologies Capital.
Momentic Momentic is a two-year-old San Francisco startup that automates end-to-end testing of web applications using AI. The company raised $3.7?million in a seed round with support from investors like Y Combinator.
Needle Operating in San Francisco and Berlin, this one-year-old startup offers an AI-powered platform for accessing and managing internal information. Needle secured $2.2?million in a seed round co-led by 468 Capital.
ReflectionAI A one-year-old Brooklyn startup founded by former Google DeepMind researchers, ReflectionAI focuses on developing autonomous AI agents. It has raised $130?million across seed and Series A rounds.?
Skyfora Skyfora, a six-year-old Helsinki startup, offers AI-powered weather forecasting by converting existing cell towers into high-resolution sensors. It raised $4.3?million in a funding round co-led by Ugly Duckling Ventures.
Viam Based in New York and five years old, Viam enables companies to apply AI to real-time data from robots, sensors, and machines without deep technical expertise. It raised $30?million in a Series C round, with total funding reaching $117?million.
VIE Technologies A nine-year-old startup from San Diego, VIE Technologie, uses AI to monitor the health of industrial equipment and predict failures. It raised $15?million in a Series A round led by Energy Impact Partners.
welevel welevel is a four-year-old Munich startup that develops AI-driven survival and city-building games with dynamic, personalized gameplay. It secured $5.7?million in a seed round led by BITKRAFT Ventures.
Zeitview An 11-year-old Los Angeles company, Zeitview uses AI and drones to inspect critical infrastructure such as solar farms and wind turbines. It raised $60?million in a funding round led by Climate Investment.
Thank you
Thank you to Bob Stefanski for contributing articles during the week.
That's all from last week,
#eLabAIReport #ai #artificialintelligence #deeptech #AIreport #zerocostAI eLab Ventures