#17 AI's $600B Question: A Time Lag, Not a Bubble
Plus: the Expectation-Reality Divide, AI in Education, Evolution of Board Portal Software, AI Books, and Notable AI News
👋 Welcome back to "AI Simplified for Leaders," your bi-weekly digest aimed at making sense of artificial intelligence for business leaders, board members, and investors.
In this issue, I cover:
Notable News and Reports: OpenAI’s AGI progress tracking framework, AI standards deep-dive, Google’s GHG emission surged, Ashurst’s AI journey
AI's $600B Question: A Time Lag, Not a Bubble
The Expectation-Reality Divide: What's Holding Gen AI Back?
AI in Education
Directors’ Corner: The Evolution of Board Portal Software
Some AI Books
I invite you to check out past issues through this link. Enjoy.
Notable AI News and Reports
1. OpenAI’s Five-Level Framework to Track Progress
OpenAI has introduced a five-level framework to track its progress towards developing artificial general intelligence (AGI). The levels are as follows:
OpenAI currently positions itself at Level 1 and is close to achieving Level 2. It is slower than what people have perceived. Meanwhile, Reuters reports OpenAI working on new reasoning technology under code name ‘Strawberry’.
2. Intuit Lays Off 1,800 Workers to Hire 1,800 in AI
In a poorly worded memo, Intuit’s CEO announced the company’s plan to lay off 1,800 workers, or 10% of its employee base, due to investments in AI. They plan to instead hire 1,800 people in engineering, product and customer-facing roles. This is the first time a major company links its layoff decisions directly to AI while not demonstrating efforts to encourage employees to re-skill or up-skill for these AI related jobs.
3. AI Standards Deep-Dive
Credo AI and Responsible AI Institute recently co-authored an article “AI Standards Deep-Dive: Decoding Different AI Standards and the EU’s Approach - Responsible AI”. The article explores the difference between standards that provide organizational level guidance, and those that include product level requirements or controls. This distinction will help organizations understand how standards development processes differ globally and the difference between existing standards.
4. Google’s GHG Emission Surged 48% Over Past Five Years
Google’s greenhouse gas emission surged 48% over the past five years due to expansion of data centers and demand from AI. Its parent company Alphabet has ended its mass purchase of cheap carbon offsets and thus stopped claiming that its operations are carbon neutral, the first time since it became carbon neutral in 2007. The company now aims to reach net-zero carbon emissions by 2030.
5. AI Fundings and Acquisitions
Investments in AI startups have more than doubled in the second quarter of 2024, reaching $24 billion. Many AI startups experience crowded competition driven by the accessibility of AI developer tools and improvements in general-purpose AI models. The valuations of some companies who get funded are very high, leading to questionable future returns for the investors.
Hebbia, a startup that uses generative AI to search large documents and respond to large questions with $13 million in revenue, has raised a $130 million Series B at a roughly $700 million valuation led by Andreessen Horowitz. The valuation surprised me given the number of capable competitors in this space.
AMD acquired Silo AI, the largest private AI lab in Europe, for $665m in cash. Silo AI specializes in developing tailored AI models, platforms and solutions for leading enterprises. The acquisition reminds me of Databricks’ acquisition of Mosaic ML.
6. A Global Law Firm Ashurst Detailed its AI Journey
A global law firm Ashurst detailed five findings from their three global GenAI trials involving 411 partners, lawyers and staff representing all of their practice areas and business services functions across 23 offices in 14 countries between November 2023 and March 2024. Particularly insightful on how they design the trials, evaluate the quality of GenAI outputs, and consider the nuances in adopting AI when its 'jagged frontier' in legal services is yet to be clarified. It is also a report full of human wisdom around the future of human-AI co-intelligence - a pleasure to read.
AI's $600B Question: A Time Lag, Not a Bubble
Is the AI gold rush headed for a spectacular crash, or are we witnessing the calm before a storm of innovation? A recent Sequoia Capital essay by David Cahn highlights a staggering $600 billion gap between projected AI revenues and investments in infrastructure, igniting debates about a potential AI bubble. However, as board directors and business leaders, it's crucial to look beyond the alarming headlines. I argue that this gap represents not an impending burst, but a natural consequence of AI's rapid evolution and the inherent lag before its wider business adoption.
Understanding the Gap
David Cahn calculates this gap by taking Nvidia's run-rate revenue forecast, doubling it to account for total AI data center costs, and doubling again to reflect a 50% gross margin for end-users of GPUs. The resulting $600 billion figure, up from $200 billion just nine months ago, seems to suggest a potential bubble in AI investments. The analysis unfortunately is less than robust and goes against the typical long-term business cycles in investments, supply and demand across different industries throughout history.
Reframing the Narrative: A Timing Lag, Not a Bubble
I propose we view this not as a problematic gap, but as a natural timing lag between infrastructure investments and revenue realization. Here's why:
Unprecedented Adoption Speed: Gen AI adoption in enterprises is occurring at one of the fastest rates we've seen for any new technology. This rapid uptake indicates strong perceived alignments to business goals and potential for future revenue growth at AI companies.
Data Readiness and Project Planning: Not all businesses can move at the same pace. Many are still in the critical stages of data preparation, experimentation, and proof-of-concept projects. The shift from experimentation to full-scale production and commercialization requires careful planning and budgeting, which naturally takes time.
Holistic Value Creation: AI's impact extends beyond direct revenue to infrastructure players. It creates value through productivity gains, cost savings, and improved outcomes across various sectors. For instance, in manufacturing, AI-powered predictive maintenance can significantly reduce downtime and extend equipment life. In finance, AI algorithms can enhance fraud detection, potentially saving billions in prevented losses. These benefits may not immediately reflect in revenue figures but contribute significantly to overall business value.
Some Evidence of Momentum
Several indicators suggest we're on the cusp of significant AI-driven business activity:
Leading AI Company Revenue Growth: OpenAI sees its annualized revenue doubled to $3.4bn over the last six months.
Cloud Revenue Growth: Major cloud service providers have become the natural choice for businesses to host their AI applications thanks to their infrastructure integration with privacy and security guardrails. For instance, Amazon AWS saw revenue surge by 17% in Q1 2024, reaching an annualized run rate of $100 billion, largely driven by generative AI workloads. Analysts project that AI-related workloads will represent 21% of AWS revenue in 2025, potentially reaching $26.9 billion.
Consulting Firm Bookings: Major consulting firms are seeing billions in new AI-related bookings. As consultants are helping businesses to establish AI strategies and implementations, the bookings already in the billions point to big companies’ commitment to AI uses in the future. I discussed the phenomenon in past newsletters and a recent New York Times report provides more numbers. Some examples:
Accenture has booked over $1.6 billion in AI-related business this year.
Boston Consulting Group expects AI consulting to contribute 20% of its revenues this year, which translates to $2.4 billion.
IBM has secured over $1 billion in sales commitments linked to Gen AI.
McKinsey anticipates that 40% of its business this year will be tied to generative AI, which would be a substantial portion of its overall revenue.
Talent Acquisition: The hiring of AI-related talent is heating up across industries. LinkedIn data shows a 21% year-over-year increase in AI-related job postings as of May 2024, indicating that companies are gearing up for significant AI initiatives.
Cross-Sector Impact: In healthcare, for example, AI is creating value in ways not directly tied to infrastructure revenue. Kaiser Permanente's Advance Alert Monitor program uses AI to prevent hospital emergencies, improving patient outcomes and reducing costs. This illustrates how AI can drive significant benefits that may not immediately appear in revenue figures.
A Summer Lull, Not a Bubble
Instead of viewing the current situation as a bubble about to burst, I see it as a summer lull – a necessary pause as businesses digest their initial AI experiments and prepare for more substantial implementations. This lull is a crucial phase where companies are laying the groundwork for future AI-driven growth.
While the outlook is promising, it's important to acknowledge potential hurdles. Data privacy concerns, ethical considerations, and the need for robust governance frameworks present ongoing challenges. Additionally, the shortage of AI talent and the complexity of integrating AI into existing business processes may slow adoption in some sectors.
Strategic Implications for Leaders
Patient Capital Allocation: Understand that the returns on AI investments may take time to materialize. The current gap represents the natural lag between infrastructure build-out and widespread commercial application.
Focus on Readiness: Use this period to ensure your organization's data infrastructure, talent, and processes are prepared for AI integration.
Look Beyond Direct Revenue: When evaluating AI initiatives, consider the full spectrum of value creation, including cost savings, improved decision-making, and enhanced customer experiences.
Strategic Partnerships: Explore collaborations with AI service providers, cloud platforms, and industry-specific AI companies to accelerate your AI capabilities without necessarily building extensive in-house infrastructure.
As for AI investments, many will lose money on overvalued business ideas with moats that turn out to be undependable. It happens in every technological investment cycle and should not be a surprise. However calling AI a bubble about to burst based on a overly simplistic calculation of $600 billion gap between AI infrastructure investments and current revenues is taking it too far. As business leaders, by focusing on readiness, strategic partnerships, and holistic value creation during this summer lull, you can ensure your organization is primed to capture the immense value AI promises to deliver, without losing your conviction amid near-term volatilities.
The Expectation-Reality Divide: What's Holding Gen AI Back?
Recently some business leaders suggest that generative AI tools may be underperforming expectations. And this contributes to the ‘summer lull’ in the section above, when business leaders are trying to diagnose the why behind the disappointment. Coding assistants are one of the areas AI seem to underperform expectations, based on software engineer discussion forums and echoed by a business leader at a recent event I attended:
"We are surprised that our engineering team only sees around 10-15% of productivity gain by using Gen AI coding copilots, much lower than the widely expected 20-30%."
At the VentureBeat Transform 2024 Conference last week, multiple attendees reported a 'meh' reaction to using Microsoft 365 Copilot, with one CEO of an enterprise AI SaaS company noting:
"If a company has more than 50% utilization of the Microsoft 365 Copilot, they are probably among the top 10% of companies we've talked to."
These statements indicate a significant discrepancy between the expected and realized benefits of Gen AI tools. While the potential of Gen AI is undeniable, several factors contribute to this underperformance:
Data Quality: Gen AI systems are only as good as the data supporting them. In many organizations, data may be siloed, inconsistent, or of poor quality, limiting the effectiveness of Gen AI tools.
Organizational Structure: Traditional hierarchies and workflows may not be optimized for Gen AI integration. Successful Gen AI adoption often requires rethinking organizational processes and decision-making structures.
Skills Gap: Many employees lack the necessary skills to effectively leverage Gen AI tools. The rapid evolution of Gen AI technology demands continuous skill development and adaptation.
Integration Challenges: Implementing Gen AI tools into existing workflows and systems can be complex and time-consuming.
Unrealistic Expectations: The hype surrounding Gen AI may have led to inflated expectations. A more realistic view of Gen AI's capabilities and limitations, grounded by a company’s strategic goals and available resources, is crucial for successful implementation.
To bridge this gap between expectations and reality, organizations need to focus on:
Improving data quality and accessibility across the organization
Adapting organizational structures to support Gen AI integration
Investing in employee training and skill development
Developing a clear Gen AI strategy aligned with business objectives
Setting realistic expectations based on industry benchmarks and pilot results
In particular, two ideas come up frequently from organizations that have seen tangible successes in their AI implementations:
Establish 'Centers of Excellence': Create dedicated teams within different business functions where employees are incentivized to deepen their Gen AI expertise specific to their roles. These teams serve as internal resources, supporting their colleagues in effectively leveraging Gen AI tools. This approach ensures that Gen AI knowledge is contextualized within each function, making it more relevant and applicable.
Form an AI Council with Change Management Capabilities: Establish a cross-functional AI council with representatives from various business units, headed by a C-level executive—preferably the Chief AI Officer or even the Chief Executive Officer. The inclusion of change management skills is vital for guiding the organization through cultural and operational shifts required for successful Gen AI adoption.
Many leaders emphasize that involvement from the top of the decision chain is paramount. It shows a strong commitment to Gen AI adoption, which is crucial when facing inevitable challenges. Without this visible support and effective change management, there's a risk that people will dismiss Gen AI tools at the first sign of performance below expectations.
By addressing these factors and implementing strong organizational support structures, businesses can work towards realizing the full potential of Gen AI tools and achieving the productivity gains they promise.
AI in Education
Many of us wear multiple hats - as business leaders, parents, grandparents, or mentors. Recently, we've all been grappling with how AI is reshaping education.
A New Book
A new book, "Teaching with AI: A Practical Guide to a New Era of Human Learning" by José Antonio Bowen and C. Edward Watson, aims to help educators navigate this new landscape.
The book covers AI basics, its impact on work and education, and strategies for integrating AI into teaching. It provides a comprehensive guide for educators to harness AI effectively, addressing issues like academic integrity and cheating. It could also satisfy your curiosity as a behind-the-scene sneak peak of what’s to come in more schools.
A Critical Perspective
But is AI really the solution to our educational challenges? Rob Nelson, Executive Director of Academic Technology and Planning at the University of Pennsylvania, offers a more skeptical view in his review of the book on the “Educating AI” Substack. When we think about AI's role in education and beyond, there are many issues we have to consider in addition to recognizing its potentials.
1. The Real Problem Isn't Technological
Nelson points out a crucial insight from the book: "Cheating is often a symptom that students do not understand or value the reward of doing the work themselves." This suggests our focus on AI might be misplaced. The core issue? Student motivation.
2. Engagement is Key
Just as in business, where we strive to help employees see the value in their work, education needs to better address the 'why' behind assignments. Students, like our employees, need to see the relevance of what they're doing.
3. Human Skills are Irreplaceable
AI can't replace human critical thinking. As Nelson notes, "Learning is hard. Generative AI is easy." This echoes what we see in business - AI tools are helpful, but they can't replace effective leadership and mentorship.
5. Liberal Arts Remain Crucial
Despite the tech focus, the book argues that "a complete education in the liberal arts has never mattered more." This underscores the enduring value of critical thinking and cultural understanding. It remains to be seen whether recruiters and hiring managers will adopt to this mindset.
How Do You Teach Your School-Age Kids About AI?
This has become a popular question lately for speakers with school-age children. Here are a few ideas to get you started:
Create a custom GPT using the OpenAI ChatGPT interface. Let kids choose a topic they're passionate about, like Taylor Swift's songs or Roblox games, and guide them through training the AI to answer questions on that topic.
Develop habits of co-creating with AI. Teach kids to view AI as a tool for enhancement, not replacement of human thinking. Encourage them to start with their own ideas before consulting AI, then use it to expand their thoughts or gain new perspectives. Emphasize the importance of questioning and verifying AI-generated information.
Talk to them about AI ethics and responsible use of AI. Introduce simple concepts like AI bias, deepfakes, frauds, and privacy using age-appropriate examples. Establish clear do's and don'ts for AI use.
Directors’ Corner: The Evolution of Board Portal Software
The evolution of board portal software has been significant over the past decade. The digitization offered a centralized platform for document sharing, collaboration, and communication, simplifying corporate governance and streamlining decision-making. User-friendly design and advanced features set the stage for further innovation.
In this “AI in the Boardroom Whitepaper” published on Nasdaq, the authors highlights that AI integration is revolutionizing board portal software. Key areas of improvement include:
Document Management: AI algorithms categorize, tag, and organize documents, making them easily searchable and accessible. It can also personalize document recommendations based on board member preferences.
Smart Agendas: AI-driven software can dynamically adapt to create agendas based on emerging trends, market dynamics, and regulatory changes.
Executive Summaries: AI can extract key information from documents, summarize reports, identify actionable insights, and pinpoint potential concerns.
Tailored Insights and Education: AI-driven personalization can tailor board members' experiences based on their expertise, preferences, and roles, suggesting resources for education and development.
One More Thing: New Books on AI
Financial Times recently reviewed three books on artificial intelligence with a common emphasis on humanity.
"How AI Thinks" by Nigel Toon: A comprehensive introduction to AI's evolution, current applications, and potential future uses. Toon explains the technology's rapid development but is less confident in addressing regulatory and policy concerns. Toon is the co-founder of British AI-model-chip startup Graphcore.
"AI Needs You" by Verity Harding: Harding examines past regulation of important technologies to guide AI control. She emphasizes the importance of political leadership, international cooperation, and clear moral and legal frameworks in fostering innovation and responsible AI development.
"As If Human" by Nigel Shadbolt and Roger Hampson: The authors explore AI ethics, advocating for treating machines as if humans were attached and holding them to high accountability standards. They propose new institutions for data management and stress the need for human involvement in AI-related decisions.
The authors emphasize the importance of human involvement in AI governance, ethical considerations, and the need for transparency and accountability. While acknowledging AI's benefits, they warn against unchecked corporate power and stress the value of human intelligence and creativity in shaping AI's future.
Have a great week.
Joyce Li