This Week In AI Business #53
This is the first week of 2025, yet a quick reminder of the news that closed 2024, as Sundar Pichai, CEO of Google, asked his executive team (referring to ChatGPT’s consumer domination): “What’s our plan to combat this in the upcoming year? Or are we not focusing as much on consumer-facing LLM?”
Do you like the weekly format? Subscribe below.
You can see what key news and trends closed 2024:
I’ve also published The Web²: The AI Supercycle, which is all you got to know, to understand where we’re going in the coming decades, in terms of development of the AI industry.
The strength of Google has always been its bottom-up consumer traction. Without it, there would be no Google.
That’s why Google, as a consumer-first company, is aware that not having an AI product as strong as ChatGPT would disrupt its core search business model.
I’ve explained this quite in detail in these issues:
That’s why in an internal strategic meeting, in one comment read aloud by Pichai, he suggested that ChatGPT “is becoming synonymous to AI the same way Google is to search,” and with that, it sounded an alarm that resonated the whole 2025, as the most critical year for Google.
If Google doesn’t aggressively gain market shares by this year to ChatGPT, with an AI native product detached from search, Google might have lost the “AI Chatbots” train for good.
And that might cost Google hundreds of billions!
The Digital Advertising cake is too good to let go
Digital advertising has become a larger market opportunity than TV.
And AI is set to transform advertising in 2025, with platforms like OpenAI and Microsoft leveraging precise targeting to capture a growing market.
Global ad spending surpassed $900 billion in 2023, with digital dominating 70% of budgets.
AI could claim 20% of the market by 2029.
The fact that the Internet has come full circle shows in how YouTube has become synonymous with TV, of course with a significant twist, which makes it a massive reach machine, a level of scale television has never experienced:
Indeed, YouTube is a cash machine within Google (Alphabet); generating over $50 billion a year, of which 70% from ads.
YouTube is the platform that most successfully managed to replicate TikTok short video format stickiness, with YouTube’s Shorts experiencing 70 billion views daily.
That’s good news for Alphabet.
Let’s go back to AI capabilities: where are we, and where are we going next?
AI moving beyond language
Up till now, AI has been primarily about language via LLMs.
Yet AI is now evolving beyond language, moving into diverse domains like robotics, vision, and autonomous problem-solving.
This is a process of tokenizing audio, video, and real-world inputs to leverage the same architecture used for language processing via transformers while building an agentic architecture on top.
This shift from generalized AI to a network of specialized, multi-modal systems mirrors human progress—from generalists to specialized generalists and experts. At the end of this process, it will be possible for companies to leverage a marketplace of AIs for specific business functions.
The state of AI Reasoning
OpenAI’s o3 model marks a significant advancement in AI, achieving state-of-the-art reasoning with innovations like program synthesis and executable Chains of Thought.
However, its high computational costs raise scalability concerns (for now).
While promising transformative potential for industries, the paradox of the AI reasoning side is that it might be an enterprise-driven industry (at least in the short term).
Indeed, only a few major enterprises might have the resources to opt for the top-tier reasoning model.
More businesses might be able to opt for the cost-effective o3-mini or cheaper AI reasoning models once they are released.
Thus, the key milestone in 2025 will be for the AI players to lower substantially the cost of AI reasoning.
The upcoming “o3-mini” might offer a cost-effective alternative for experimentation once released, together with competitors, to drive the cost of AI reasoning (“system two thinking”) down substantially.
Yet again, this is all to see.
The AI Enterprise Push
At an enterprise level, the pressure to integrate AI into the company's core has become a top-down prioritization effort amid increased competition across many sectors.
That has already caused the rise of the “AI Managerial Class.”
In fact, for one thing, AI has caused a surge in AI-focused Leadership Roles as companies are heavily investing in AI-related C-suite roles, which spiked by 428% in two years, with VP, director, and manager roles also seeing triple-digit growth.
Why does that matter? Well, as I’ve explained in my predictions for 2025:
The counterintuitive take here is that, contrary to many other industries, this AI paradigm started with a massive consumer push (with the exposition of AI Chatbots), and only later did we get Enterprise running after it to understand how to adopt the technology.
In this catch-up game, enterprises are experiencing “AI anxiety” across the whole of 2025 as they’ll try to figure out how, where, and what to integrate within their core.
Enterprise AI Trends to Expect in 2025
TechCrunch has amassed an impressive survey of Enterprise AI trends for 2025, according to many vertical executives. I’ve summarized them in the graphic above.
Yet, to me, the Top Five are:
AI Adoption and Data Quality: AI adoption will hinge on better data quality as enterprises move from experimentation to large-scale deployment.
Code Agents and App Modernization: AI will increasingly modernize mainframe apps, re-platform older codebases, and improve app development efficiency.
Automation in High OpEx Sectors: AI might enable automation in traditionally untouchable sectors like accounting services, legal workflows, and revenue cycle management, achieving SaaS-like margins.
Enterprise Sales Cycles and Pricing Models: Understanding the duration of enterprise tool trials and exploring new pricing models (e.g., SaaS, consumption-based, outcome-based) will be key.
Time-to-First-Value (TTFV): TTFV will become a critical metric for evaluating measurable benefits to customers after adoption. As enterprises prioritize ease of implementation and ROI, solutions with shorter TTFVs gain an edge in adoption, showing immediate value in streamlining workflows and reducing operational friction.
Since I’ve referenced Enterprise trends, let’s look at the player building its strength on the enterprise side.
Microsoft’s going all in on AI Infrastructure
There is no better news to frame 2025 than the just-announced plan of Microsoft to double down on AI Infrastructure as it plans to invest $80 billion in fiscal 2025 to build AI-focused data centers, with over half of the spending in the U.S.
This investment supports partnerships with OpenAI and others to boost further Azure, which experienced 33% year-over-year growth.
Microsoft emphasizes the need for U.S. AI leadership amid growing global competition, particularly from China.
This news is critical as it highlights a few key things:
AI Hardware Bottleneck: As I’ve highlighted many times, we are nowhere near the infrastructure needed to serve the AI demand so far. Microsoft is well aware of it. And while this massive spending might turn out to be an “overshooting.” The reality is, so far, overshooting, at this stage, might be more rational than undershooting. If AI adoption consolidates into the 2030s, any foundational company position might gain trillions in market cap.
AI Race is quite profitable for now: of course, this overshooting also comes from the massive competition among hyperscalers on who has the most extensive infrastructure to serve the gigantic inference demand. The players able to serve the market at the inference level might also experience healthy and great gross margins (the inference cost for these big tech players is relatively low compared to the money they make out of it). Unless, in 2026, we’ll have a “market reckoning” with major enterprises pulling back most of the budget, whoever has the inference to serve a good chunk of the AI market demand will make billions out of it.
These two reasons alone explain why what might seem an overshoot might be, for now, a rational choice.
In addition to that, Microsoft announced it with a whole manifesto, which it called “The Golden Opportunity for American AI:”
The key point here is that while we’ll see a major rise of consumer applications for AI, we’re also living a completely different geopolitical phase.
AI, as an industry, in the next decade will live in between of a tension of governments competing to get as much of infrastructural capacity within their borders, and most of the bill paid by large enterprises.
With enterprise businesses covering a good chunk of the AI costs, every major tech player will be going after it.
Alibaba has officially entered the AI Enterprise race.
As reported by CNBC, Alibaba Cloud has drastically reduced prices for its AI language model Qwen-VL by up to 85%, targeting enterprise users amid fierce competition among Chinese tech giants.
With over 90,000 deployments, this move follows earlier price cuts and reflects Alibaba’s strategy to lead China’s AI market by attracting businesses with cost-effective solutions.
But where are we in the development of the whole ecosystem?
The proof is in the substrate
To understand the massive pace of AI demand, take the substrate.
In semiconductor manufacturing, it is a thin, flat material that serves as the base layer or foundation for building electronic circuits.
Specifically, in AI chips, substrates are critical as they connect the chip to the circuit board.
Among other critical features, the substrates facilitate the transmission of electrical signals between the chip (e.g., CPU, GPU) and the larger electronic device (like a computer or AI server).
This is the key component of Nvidia’s AI GPU.
Ibiden, a key supplier of chip substrates for Nvidia’s AI semiconductors, is experiencing unprecedented demand driven by Nvidia’s AI dominance. Ibiden’s current expansion plans, including a new factory in Japan, may not meet the needs of clients like Nvidia, Intel, and AMD.
That’s where we are in terms of the gap between the demand and supply of key components for AI chips.
This is a key sign of understanding how early we are in getting the underlying AI infrastructure built to serve the pent-up demand, as key components are becoming scarce across the whole AI supply chain.
“Selling picks and shovels in the AI gold rush” generated $2 trillion in market value for NVIDIA by 2024
Last year, NVIDIA’s market valuation increased by over $2 trillion, pushed by the AI adoption wave, with AI hyperscalers spending over $200 billion on GPUs.
As the first financial week of 2025 closed on Friday, NVIDIA almost passed Apple's market capitalization, as Apple closed at nearly $3.7 trillion, while NVIDIA closed the session at over $3.5 trillion.
NVIDIA has silently turned into an AI VC Fund in 2024
Nvidia invested $1 billion by 2024 across 50 startup funding rounds, achieving a tenfold growth compared to 2022.
As reported by The Financial Times, in 2024 - a period where VCs outside the major ones are struggling to deploy capital successfully - NVIDIA has acted as the most effective firm.
I’ve recapped below the whole strategy that NVIDIA has used to deploy a billion dollars in the AI space:
Notable acquisitions included Run.ai, Nebulon, OctoAI, Brev.dev, Shoreline.io, and Deci, focusing on AI workload management and software.
Strategic investments targeted major AI players like xAI, OpenAI, Cohere, and Mistral, alongside infrastructure companies CoreWeave ($100M) and Applied Digital ($160M).
Investment focus areas span medical technology, gaming, drones, chips, traffic management, logistics, NLP, humanoid robots, and cloud computing.
Some key patterns came out of it:
Focus on AI Dominance: Nvidia’s investments and acquisitions demonstrate a clear strategy to dominate the AI ecosystem, from foundational AI models (e.g., OpenAI, Cohere) to infrastructure (e.g., CoreWeave).
End-to-End Ecosystem Building: Investments span the entire AI pipeline, including software (Deci, xAI), hardware (chips, GPUs), and infrastructure (cloud computing).
Scaling Through Acquisitions: Acquiring specialized startups like Run.ai and Nebulon helps Nvidia scale specific AI functionalities like workload management.
Infrastructure as a Backbone: Strategic investments in CoreWeave and Applied Digital reveal a focus on building robust infrastructure for AI adoption.
NVIDIA’s massive bet on general-purpose robotics is inaugurating the “year of the humanoid”
As reported by The Financial Times, Nvidia is pivoting to robotics as a key growth driver ahead of rising competition in AI chips.
It’s launching Jetson Thor in 2025 and investing in simulation-based training to bridge the “Sim-to-Real gap.”
With tools spanning AI training to robotic hardware, Nvidia aims to lead the projected $165B robotics market by 2029.
Why, how, and what challenges can we expect there?
Nvidia aims to lead the imminent robotics boom, launching its Jetson Thor compact computers developed specifically for humanoid robots in 2025.
The “ChatGPT Moment” for Robotics might be coming as Nvidia’s VP, Deepu Talla, believes robotics is at a “tipping point,” driven by generative AI and simulation-based training advancements.
That’s why NVIDIA’s Full Stack Robotics Solution is coming via AI training, simulation environments in Omniverse, and chips as robotic “brains.”
The company also placed a massive strategic investment in Figure AI, a humanoid robotics company, at a $2.6 billion valuation to accelerate robotics advancements.
A key challenge right now is overcoming the “Sim-to-Real gap.” This problem states the policies trained in simulations don’t necessarily translate/ transfer to real-world environments.
This challenge arises due to the “reality gap,” which includes discrepancies in dynamics, sensory inputs, and physical behaviors between simulated and real settings. This might prevent robots from operating effectively in real-world settings.
NVIDIA’s tools try to address exactly that.
Customers like Amazon, Toyota, and Boston Dynamics already leverage Nvidia’s robotics tools for warehouse automation and autonomous systems.
The global robotics market, valued at $78 billion in 2024, is projected to double to $165 billion by 2029, creating substantial growth opportunities.
This is how NVIDIA is tackling it:
Indeed, NVIDIA's multi-layered approach to AI robotics also shows how the company is betting big on it as the next demand booster.
1. AI Models (Top Layer)
2. Development Tools (Middle Layer)
3. Hardware Layer (Bottom Layer)
This layered approach allows NVIDIA to support the entire lifecycle of robotics development, from conceptual AI modeling to real-world implementation.
That’s how NVIDIA got ready for real-world AI development's next milestone.
Samsung also doubled down on humanoid robots
As NVIDIA is pushing the boundaries of AI infrastructure for humanoids, Samsung is following suit by increasing its stake in South Korea’s Rainbow Robotics to 35% for $181 million, making it the largest shareholder.
Rainbow Robotics will become a subsidiary, accelerating Samsung’s development of humanoid robots.
A new Future Robotics Office will oversee initiatives, positioning Samsung alongside Tesla, Nvidia, and Microsoft in the robotics race.
What’s coming next?
CES 2025 Is coming on January 7-10, with over 4,300 companies and 140,000 attendees expected.
What key trends to expect?
AI Everywhere: AI is featured in everything from cars to refrigerators, with some applications solving real problems while others remain speculative.
Robotics: Expanded focus on manufacturing, transportation, and humanoid robots from companies like Hyundai and Toyota.
Automotive Tech: CES solidifies its place as a leading automotive show with participation from Sony, Toyota, and Hyundai.
Smart Home Revival: New devices leveraging the Matter standard, generative AI, and updates from Google, Amazon, and Apple.
Extended Reality Challenges: Companies like Qualcomm are pushing forward despite Vision Pro and Quest’s mixed market performance.
Recap: In This Issue!
Google’s AI Challenges
Google recognizes ChatGPT as a major competitor, likening its AI dominance to Google's in search.
Sundar Pichai emphasized the urgency to develop an AI-native product separate from search to maintain market leadership.
Failure to catch up in AI chatbots could cost Google hundreds of billions.
Digital Advertising Trends
Digital advertising surpassed $900 billion in 2023, with 70% allocated to digital platforms.
AI is projected to transform advertising, capturing 20% of the market by 2029.
Platforms like OpenAI and Microsoft leverage AI for precise targeting, reshaping the market.
AI Advancements Beyond Language
AI is transitioning from language-based models to multi-modal systems (e.g., robotics, vision).
Specialized AI systems are emerging, enabling targeted applications across industries.
This shift could lead to a marketplace of AIs for diverse business functions.
AI Reasoning and Scalability
OpenAI’s o3 model showcases advanced reasoning but faces scalability challenges due to high costs.
Cost-effective alternatives like “o3-mini” aim to democratize AI reasoning capabilities.
Reducing the cost of reasoning models will be a major milestone in 2025.
Enterprise AI Push
Enterprises are prioritizing AI integration, leading to a surge in AI-focused leadership roles (428% growth in two years).
Key trends for 2025:
Improved data quality for AI adoption.
Modernization of legacy apps through AI.
Automation in high OpEx sectors.
Shorter time-to-first-value for enterprise AI solutions.
Microsoft’s AI Infrastructure Investments
Microsoft plans $80 billion in AI-focused data center investments for fiscal 2025.
Emphasis on addressing AI hardware bottlenecks and ensuring U.S. AI leadership.
Investments position Microsoft as a leader in serving AI inference demand.
Nvidia’s Dominance in AI
Nvidia’s market value surged by $2 trillion in 2024, driven by AI chip demand.
Investments in startups and robotics signal a strategic pivot toward long-term AI ecosystem dominance.
Robotics, supported by tools like Jetson Thor, is a key growth driver for Nvidia.
Robotics and Humanoids
Nvidia and Samsung are advancing in robotics, addressing challenges like the "Sim-to-Real gap."
Global robotics market expected to grow from $78 billion (2024) to $165 billion (2029).
CES 2025 to highlight innovations in robotics and AI integration across industries.
AI Hardware Bottlenecks
Supply chain constraints for critical components (e.g., chip substrates) underscore the early stage of AI infrastructure development.
Companies like Nvidia and Ibiden are scaling production to meet unprecedented AI demand.
Upcoming CES 2025 Trends
AI integration across consumer tech (e.g., cars, smart homes, extended reality).
Focus on robotics and automotive advancements.
Innovations from companies like Google, Amazon, and Apple in generative AI applications.
Keep an eye on these!
With massive ♥️ Gennaro Cuofano, The Business Engineer
It’s all pointing to a new version of the “space race” with AI. What are your thoughts on talent acquisition in that area? Every player will of course want top talent, but what will that mean for international hires and a plausible conflict of interest?