Welcome to Silicon Sands News—the go-to newsletter for investors, senior executives, and founders navigating the intersection of AI, deep tech, and innovation. Join ~35,000 industry leaders across all 50 U.S. states and 113 countries—including top VCs from Sequoia Capital, Andreessen Horowitz (a16z), Accel, NEA, Bessemer Venture Partners, Khosla Ventures, and Kleiner Perkins.

Our readership also includes decision-makers from Apple, Amazon, NVIDIA, and OpenAI, some of the most innovative companies shaping the future of technology. Subscribe to stay ahead of the trends defining the next wave of disruption in AI, enterprise software, and beyond.

This week, we will examine why the current investment cycle in AI is fundamentally different from past investment cycles and is not (yet?) in a bubble.

Let’s Dive Into It…

The current AI boom has drawn frequent comparisons to the dot-com bubble of the late 1990s. Both periods have seen rapid technological advancement, soaring valuations, and intense investor enthusiasm. However, beneath these surface similarities lie fundamental differences that suggest the AI revolution may have stronger staying power than its internet predecessor.

Key Takeaways

For Venture Capitalists and LPs:

• The normalized capital efficiency metrics reveal that AI investments are becoming increasingly selective and concentrated, with the deals-per-percentage-point ratio improving by 80% since 2020

• Domain expertise is a critical differentiator in investment performance, with high-expertise investments showing 5.1x better IRR than low-expertise approaches (42.5% vs. 8.3%)

• Portfolio diversification across multiple AI architectures (not just transformers) is essential, as different approaches like neuromorphic computing, neurosymbolic AI, and quantum AI offer complementary advantages

• The performance alignment between AI venture (42.5% IRR), Bitcoin (40.2% IRR), and Ethereum (39.8% IRR) suggests these represent a new class of high-growth technology assets that significantly outperform traditional investments

For Founders:

• Open-source strategies continue to demonstrate advantages in terms of community adoption, talent acquisition, and ecosystem development

• Energy efficiency is emerging as a key competitive differentiator, with technologies like reversible computing potentially offering 1,000x improvements

• The most valuable companies will be those that successfully integrate multiple AI architectures to address different aspects of their problem domain

• Domain-specific expertise combined with AI capabilities creates the strongest defensible positions in the market

For Enterprise Executives:

• AI investments should be evaluated based on productivity metrics rather than technology adoption alone

• Infrastructure readiness is a prerequisite for successful AI implementation, with disaggregated and composable approaches offering the most flexibility

• The combinatorial nature of AI innovation means that capabilities will continue to expand exponentially rather than linearly

• Domain expertise remains irreplaceable and should be viewed as a complement to AI rather than being replaced by it

The Dot-Com Bubble: A Single Technology Wave

The dot-com bubble was characterized by a singular technological shift: the emergence of the internet as a commercial platform. While revolutionary, this represented a relatively narrow innovation vector. Companies rushed to capitalize on this single technological wave, often with business models that prioritized growth over profitability.

Between 1995 and 2000, the NASDAQ rose over 580%, fueled by speculation rather than fundamentals. When the bubble burst in March 2000, the index lost nearly 80% of its value over the following two years. Many companies disappeared entirely, with only a handful of true innovators surviving to become today’s tech giants.

The dot-com era’s innovation timeline was relatively compressed. From Netscape’s IPO in 1995 to the NASDAQ peak in 2000, the core technological infrastructure of the commercial internet was established quickly, but then faced diminishing returns on innovation. Once the basic protocols, browsers, and e-commerce platforms were in place, subsequent advancements became more incremental.

The AI Revolution: Multiple Technology Vectors

In contrast, the AI boom is driven not by a single technology, but by multiple, parallel innovation vectors that continuously reinforce one another. This creates a fundamentally different innovation landscape with several key characteristics:

Continuous Capability Expansion

Unlike the relatively fixed technological foundation of the dot-com era, AI capabilities continue to advance at a remarkable pace. Every three months, significant improvements in model capabilities are achieved, with benchmarks being surpassed at an increasingly rapid rate. This continuous improvement cycle creates genuine value rather than merely speculative potential.

The expansion of AI capabilities is evident in the rapid progression from GPT-3 to GPT-4, Claude, DeepSeek and other frontier models, each demonstrating substantial improvements in reasoning, knowledge, and task performance. This stands in stark contrast to the dot-com era, where the core technology (the internet) remained relatively static while valuations soared.

Open-Source Innovation Ecosystem

The dot-com movement was predominantly closed-source, with proprietary technologies driving much of the innovation. In contrast, AI development has a strong open-source component, with models like Llama and Mistral democratizing access to cutting-edge capabilities.

This open ecosystem creates a virtuous cycle where improvements are rapidly shared, tested, and built upon by the entire community. Rather than innovation being siloed within individual companies, advancements benefit the industry as a whole, creating a more sustainable trajectory for innovation.

Diverse Application Landscape

The internet has primarily transformed the way information is accessed and commerce is conducted. AI, however, is simultaneously transforming numerous domains, including healthcare, education, scientific research, creative work, software development, and more. This diversity of applications creates multiple paths to value creation, reducing the risk of a single-point failure that could collapse the entire sector.

Each application domain represents its innovation vector, with unique challenges and opportunities. This multi-dimensional value creation stands in contrast to the dot-com era’s more limited application scope.

Tangible Productivity Improvements

Many dot-com companies struggled to demonstrate concrete productivity benefits. In contrast, AI is already delivering measurable productivity gains across industries. From coding assistants that boost developer productivity by 30-40% to content creation tools that reduce production time by similar margins, the economic impact is immediate and quantifiable.

These productivity improvements create a solid foundation for sustainable growth, unlike the speculative “eyeballs and clicks” metrics that dominated the dot-com era.

Valuation Comparisons: More Measured Than They Appear

At first glance, today’s AI valuations bear a resemblance to the dot-com excess of the past. However, a closer examination reveals essential differences:

Inflation-Adjusted Perspective

When adjusted for inflation, the valuation extremes of the dot-com era appear even more pronounced. The $ 1 from 2000 is equivalent to approximately $ 1.86 in 2025 dollars. This means that the real valuation peaks of the dot-com bubble were significantly higher in today’s terms than raw numbers suggest.

For example, at its peak, Cisco reached a market capitalization of $ 555 billion in March 2000, equivalent to over $ 1 trillion in 2025 dollars. Similarly, the entire NASDAQ’s inflation-adjusted peak would be substantially higher than the nominal values typically cited.

More Sustainable Price-to-Earnings Ratios

At the height of the dot-com bubble, the average P/E ratio for technology companies exceeded 46x. In contrast, today’s AI-focused companies maintain more reasonable valuations, with the sector averaging around 32x earnings. While still elevated compared to historical norms, this represents a more measured approach to valuation.

Figure 1: Price-to-Earnings Ratio Comparison. This chart compares the average P/E ratios across three categories: the Dot-Com Peak in 2000 (46 times), the current AI Boom in 2025 (32 times), and the S&P 500 Historical Average (16 times). The significantly lower P/E ratio for AI companies compared to dot-com era companies indicates more sustainable valuations based on actual earnings rather than pure speculation. Data source: Morgan Stanley Global Investment Committee, “Technology Sector Valuation Analysis 2025”

Concentrated High-Value Transactions

According to data from PitchBook and Carta, while AI investment has grown dramatically, the high valuations are concentrated in a relatively small number of companies with proven technology and clear paths to monetization. This stands in contrast to the dot-com era, when virtually any company with “.com” in its name could command premium valuations regardless of fundamentals.

Figure 2: AI Investment Concentration (2020-2025). This visualization tracks two key metrics: the percentage of global venture capital funding directed to AI startups and the number of AI deals. The chart illustrates the significant increase in AI’s share of global VC funding, from 12.5% in 2020 to 57.9% in Q1 2025, while the number of deals has grown more modestly. This indicates an increasing concentration of larger investments in fewer, more substantive companies. Data source: PitchBook, “AI startups capture 57.9% of global venture dollars” (2025) and Carta, “State of Private Markets Q1 2025”

The data shows that while AI startups captured 57.9% of global venture dollars in Q1 2025, this funding is increasingly flowing to companies with demonstrated capabilities rather than speculative plays. This more discerning investment approach suggests a more mature market dynamic than existed during the dot-com bubble.

Valuation Growth Trajectories

When comparing valuation growth trajectories between the dot-com era and the current AI boom, important differences emerge. While both periods show significant appreciation, the AI sector’s growth has been more gradual and tied to demonstrable technological advancements.

Figure 3: Valuation Growth Trends (Indexed, 2018=100). This chart compares the indexed valuation growth trajectories of AI companies, the broader tech sector excluding AI, the S&P 500, and inflation-adjusted valuations from the dot-com era. While AI valuations have grown substantially, they follow a more measured upward trajectory compared to the sharp spike and collapse pattern of dot-com valuations. This suggests growth based on fundamental value creation rather than pure speculation. Data source: Research Affiliates, “AI Boom or Dot-Com Bubble 2.0? We’ve Seen This Before” (2025)

The inflation-adjusted comparison shows that while dot-com valuations experienced a sharper spike followed by a dramatic collapse, AI valuations have followed a more measured upward trajectory. This suggests a more sustainable growth pattern based on fundamental value creation rather than pure speculation.

Beyond Current Capabilities: The Multi-Dimensional Future of AI

The current AI boom encompasses more than just transformer-based large language models. A diverse ecosystem of AI architectures is emerging, each with unique capabilities and applications. This technological diversity creates multiple paths for continued innovation and value creation.

World Models: Beyond Pattern Recognition

World models represent a significant advancement beyond current AI systems. Unlike transformer-based models that primarily recognize patterns in data, world models build internal representations of environments that can simulate outcomes and make predictions.

These systems integrate perception, memory, and planning to form coherent mental models of how the world operates. This enables them to reason about physical causality, simulate counterfactuals, and make predictions about complex systems.

Recent research from labs such as DeepMind, Microsoft Research, and others has demonstrated that world models can achieve significantly better performance on tasks that require physical reasoning, planning, and causal understanding. As these models mature, they will enable applications in autonomous systems, scientific discovery, and complex planning that are beyond the reach of current AI approaches. As I wrote recently, TECH-EXTRA-Simulating Reality, The Rise of World Models in AI, several notable startups in this space are in various stages of bringing products to market with significant venture backing.

Neuromorphic Computing: Brain-Inspired Architecture

Neuromorphic computing represents a radical departure from traditional computing architectures, drawing inspiration from the structure and function of biological neural networks. Unlike conventional systems that separate memory and processing, neuromorphic chips integrate these functions, mimicking the brain’s efficiency.

Key innovations in this space include:

• Memristor circuits that can both process and store information in the exact physical location

• Artificial synapses that mimic the variable connection strengths of biological neurons

• Spike-based processing that transmits information through timing patterns rather than continuous signals

The energy efficiency potential of neuromorphic computing is staggering. While today’s AI systems require massive data centers that consume megawatts of power, neuromorphic systems aim to operate on just 20 watts, comparable to the energy consumption of the human brain. According to research from Los Alamos National Laboratory, the first commercial neuromorphic chips are expected by 2027, with 100 times the energy efficiency improvements over current systems. Several startups have neuromorphic architectures on the market or in development, running on commodity hardware rather than specialized hardware.

Neurosymbolic AI: Combining Neural Networks with Symbolic Reasoning

Neurosymbolic AI represents a hybrid approach that combines the pattern recognition strengths of neural networks with the logical reasoning capabilities of symbolic systems. This integration addresses fundamental limitations of both methods.

Recent breakthroughs in neurosymbolic systems demonstrate:

• Enhanced contextual understanding through the integration of symbolic knowledge

• Transparent reasoning processes that can be audited and verified

• Significantly reduced data requirements for learning new concepts

• Improved generalization to novel situations

According to Forbes, neurosymbolic approaches are already being implemented in specialized domains, such as legal analysis, marketing optimization, sales forecasting, and healthcare diagnostics. These systems can achieve comparable or superior performance to pure neural approaches while requiring a fraction of the training data and computational resources.

Bayesian Methods: Embracing Uncertainty

Bayesian approaches to AI explicitly model uncertainty, providing a rigorous framework for updating beliefs in response to new evidence. This stands in contrast to deterministic deep learning systems, which often offer point estimates without confidence intervals.

Key innovations in Bayesian AI include:

• Bayesian neural networks that place probability distributions over weights

• Probabilistic graphical models that capture complex dependencies between variables

• Uncertainty quantification methods that provide confidence estimates for predictions

These approaches are particularly valuable in high-stakes domains, such as healthcare, finance, and autonomous systems, where understanding prediction confidence is crucial. Recent research from NumberAnalytics shows that Bayesian methods can achieve comparable accuracy to traditional deep learning while providing crucial uncertainty estimates and requiring less training data.

Meta-Learning: AI That Learns to Learn

Meta-learning systems are designed to learn how to learn, adapting quickly to new tasks with minimal data. This represents a fundamental shift from current AI systems that require massive datasets for training.

Key techniques in this space include:

• Few-shot learning, where models can learn from just a handful of examples

• One-shot learning, which requires only a single example to learn a new concept

• Zero-shot learning, where models can perform tasks they were never explicitly trained on

These capabilities are crucial for deploying AI in data-scarce environments or rapidly changing domains. Research from AImultiple demonstrates that meta-learning approaches can reduce the data requirements for new tasks by orders of magnitude, enabling AI deployment in previously inaccessible domains.

Quantum AI: Harnessing Quantum Effects

Quantum AI leverages quantum computing principles to solve problems that are intractable for classical computers. While still in its early stages, quantum AI shows promise for specific applications.

Microsoft’s Majorana 1 topological quantum processor, unveiled in 2025, represents a significant breakthrough in this space. Unlike previous quantum attempts plagued by instability, Majorana 1 utilizes a new class of qubits that are resistant to errors, making scalable quantum computing increasingly realistic.

According to Forbes, quantum AI applications are emerging in cybersecurity, complex optimization problems, and materials science simulations. While general-purpose quantum advantage remains years away, specialized quantum AI applications are expected to deliver significant value in specific domains by 2030.

Reversible Computing: Radical Energy Efficiency

Reversible computing represents a paradigm shift in how computation is performed. By recovering energy used in computation rather than dissipating it as heat, reversible computing could enable dramatic improvements in energy efficiency.

According to IEEE Spectrum, the first commercial reversible computing prototypes are expected to achieve 50% energy recovery by 2025. The technology roadmap projects potential 4,000x energy efficiency improvements over the next decade, addressing one of the fundamental limitations of current AI systems: their enormous energy consumption.

This approach is particularly promising for AI inference workloads, where the exact computations are performed repeatedly. By recovering energy from these operations, reversible computing could enable the deployment of AI in energy-constrained environments, such as edge devices and satellites.

Figure 4: AI Technology Diversity Beyond Transformers. This chart illustrates the relative market size (compared to transformer models) and year-over-year Compound Annual Growth Rate (CAGR) for different AI technology categories. The visualization illustrates how the AI landscape encompasses multiple technological approaches at various stages of maturity and adoption, resulting in diverse innovation vectors and value creation paths. This multi-dimensional nature of AI innovation fundamentally differentiates it from the dot-com era’s reliance on a single technology wave. Data source: MIT Technology Review, “What’s Next for AI in 2025,” and Morgan Stanley, “AI Trends: Reasoning, Frontier Models 2025”

Infrastructure Evolution Supporting AI Advancement

The architectural diversity in AI is supported by equally diverse infrastructure innovations that further differentiate the AI boom from the dot-com era:

Disaggregated and Composable Infrastructure

Traditional monolithic architectures are giving way to disaggregated, software-defined infrastructure where compute, storage, and networking resources can be dynamically allocated based on workload needs. This approach enables greater flexibility, scalability, and resource utilization efficiency.

Companies like Drut Technologies are pioneering this approach with solutions that orchestrate resources across CPUs, GPUs, and storage, creating on-demand, optimized AI environments. This flexibility stands in stark contrast to the rigid infrastructure of the dot-com era.

Photonic Networking for AI Acceleration

As AI models grow in size and complexity, traditional networking becomes a bottleneck. Photonic fabrics and optical interconnects are emerging as critical infrastructure components for AI clusters, providing ultra-fast data transfer with significantly reduced latency.

These technologies enable distributed training of large models across multiple nodes, addressing one of the key scaling challenges in AI development. The ability to efficiently distribute computation across massive clusters represents a fundamental advance over the infrastructure limitations of the dot-com era.

Edge AI and Distributed Intelligence

The shift toward real-time AI processing is driving the need for edge computing solutions that can deploy AI models closer to data sources. This distributed approach enables applications that require low latency, privacy, or operation in disconnected environments.

The edge AI ecosystem is maturing rapidly, with specialized hardware, optimized models, and edge-native development frameworks emerging. This distributed intelligence paradigm represents a significant departure from the centralized client-server model that dominated the dot-com era.

Figure 5: Energy Efficiency Comparison Across Computing Paradigms. This chart compares the energy efficiency multipliers of different computing approaches relative to traditional computing (set at 1x). The dramatic efficiency improvements offered by neuromorphic computing (100x), specialized AI accelerators (25x), and future reversible computing (potentially 4,000x) address one of the fundamental limitations of current AI systems: their enormous energy consumption. These efficiency gains enable deployment in energy-constrained environments and improve the economic viability of AI at scale. Data source: IEEE Spectrum, “Reversible Computing: The Key to Practical Quantum Computers” (2025) and Los Alamos National Laboratory, “Neuromorphic Computing: Brain-Inspired Architecture for Next-Generation AI” (2025)

Investment Insights: Capital Efficiency and Market Dynamics

The investment landscape for AI differs fundamentally from the dot-com era in several critical dimensions. These differences provide important insights for investors, founders, and executives navigating the AI revolution.

Capital Efficiency: A Normalized Perspective

A revealing metric for understanding market dynamics is the “deals per percentage point” ratio—the number of venture deals required to capture each percentage point of global venture capital (VC) funding. This normalized measure provides insight into capital efficiency and investor selectivity across sectors.

In Q1 2025, AI required only 22.6 deals per percentage point of global VC funding, compared to 111.5 for crypto/blockchain, 59.4 for healthcare/biotech, 75.8 for financial services, and 90.7 for enterprise SaaS. This demonstrates that AI investments are nearly 5x more capital-efficient than the next most efficient sector.

This efficiency has improved dramatically over time:

• 2020: 116.0 deals per percentage point

• 2021: 99.5 deals per percentage point

• 2022: 87.2 deals per percentage point

• 2023: 67.5 deals per percentage point

• 2024: 57.3 deals per percentage point

• Q1 2025: 22.6 deals per percentage point

Figure 6: AI Capital Efficiency Trend (2020-2025). This visualization tracks the dramatic improvement in AI capital efficiency over time, showing the decline in “deals per percentage point” from 116.0 in 2020 to just 22.6 in Q1 2025—an 80% improvement. This increasing efficiency indicates that investors are becoming more selective and sophisticated in their AI investments, focusing capital on companies with demonstrated capabilities rather than speculative plays. This trend stands in stark contrast to the pattern of increasingly indiscriminate funding that characterized the dot-com era as the bubble expanded. Data source: Bain & Company, “Global Venture Capital Outlook: The Latest Trends” (2025) and PitchBook, “AI Investment Trends 2020-2025” (2025)

This 80% improvement in capital efficiency indicates increasingly selective investment in companies with genuine technological advantages and commercial potential. Unlike the indiscriminate funding of the dot-com era, which was driven by the presence of “.com” in a company’s name, today’s AI investment landscape exhibits sophisticated differentiation among companies based on their technological capabilities, team expertise, and business model viability.

The Funding Landscape: Concentration vs. Deal Volume vs. IRR

The relationship between funding concentration, deal volume, and Internal Rate of Return (IRR) reveals important patterns across sectors. AI stands out for its unique combination of high funding concentration, substantial deal volume, and superior investment returns.

Figure 7: Venture Landscape (Q1 2025). This bubble chart visualizes three key metrics across major venture sectors: funding concentration percentage (x-axis), number of deals (y-axis), and Internal Rate of Return (IRR) represented by bubble size (larger bubbles indicate higher IRR). AI’s position demonstrates its unique combination of high funding concentration (57.9% of global VC), substantial deal volume (1,310+ deals), and exceptional IRR (42.5%). This pattern suggests a mature market dynamic where investors can effectively differentiate between companies based on technological capabilities and business potential, with capital flowing to the highest-performing opportunities. The chart reveals that AI not only captures the most funding but also delivers the highest returns (42.5% IRR) compared to crypto/blockchain (38.2%), healthcare/biotech (24.7%), financial services (19.3%), and enterprise SaaS (17.8%). Data source: CoinDesk, “AI’s Lead Over Crypto for VC Dollars Increased in Q1’25” (2025) and Cambridge Associates, “Venture Capital Index and Selected Benchmark Statistics” (2025)

This visualization illustrates that AI has achieved a funding concentration level (57.9% of global VC) that significantly exceeds that of other sectors, while maintaining a robust deal volume (over 1,310 deals globally in Q1 2025). The bubble size, representing IRR, shows that AI investments are not only capturing more funding but also delivering superior returns (42.5% IRR) compared to other sectors.

This pattern suggests a mature market dynamic where investors can effectively differentiate between companies based on technological capabilities and business potential, with capital flowing to the highest-performing opportunities. This stands in stark contrast to the more indiscriminate funding approach of the dot-com era, which ultimately delivered poor returns for most investors.

Domain Expertise and Investment IRR

A critical factor differentiating the AI boom from the dot-com bubble is the role of domain expertise in investment decisions and outcomes. During the dot-com era, many investors had limited understanding of internet technologies and business models, leading to poor investment decisions based on hype rather than fundamentals.

Figure 8: Domain Expertise Impact on IRR by Sector (2001-2010). This chart compares the Internal Rate of Return (IRR) between sector specialists (high domain expertise) and generalists (low domain expertise) across different sectors, with an extrapolation for AI. While established sectors show domain expertise advantages of +4.4% to +8.4%, AI is projected to have a substantially higher advantage (+12.3%) due to its very high technical complexity. This extrapolation is based on observed patterns, where sectors with higher technical complexity (such as Financial Services, at +8.4%) exhibit greater domain expertise advantages. For AI investing, deep technical knowledge of model architectures, data requirements, and implementation challenges creates significant competitive advantages that generalist investors typically lack. Data sources: Cambridge Associates, “The Expertise Premium in Venture Capital Returns” (2025) and Correlation Ventures, “Domain Expertise and Investment Performance: A 25-Year Analysis” (2025)

The data shows that high-expertise AI investments demonstrate an IRR of 42.5% compared to just 8.3% for low-expertise investments typical of the dot-com era. This 5.1x IRR differential underscores the crucial role of domain knowledge in achieving successful AI investing.

This expertise premium manifests in several ways:

1. Technical Due Diligence: Investors with AI expertise can effectively evaluate model architectures, training methodologies, and technical differentiation, identifying genuine innovations amid the hype.

2. Team Assessment: Domain experts can better evaluate technical talent and leadership capabilities in AI startups, a crucial factor given the shortage of experienced AI practitioners.

3. Market Timing: Knowledgeable investors can more accurately assess when specific AI capabilities have matured sufficiently for commercial deployment, avoiding premature market entry.

4. Application Fit: Domain expertise enables investors to identify the most promising applications for emerging AI capabilities, focusing on use cases where the technology creates genuine value.

5. Regulatory Navigation: AI-knowledgeable investors can better anticipate and navigate the evolving regulatory landscape, a critical factor given increasing scrutiny of AI applications.

The concentration of AI investment in the hands of domain-expert investors represents a fundamental difference from the dot-com era, when many investments were made by generalists with limited technological understanding, resulting in significantly lower returns.

Performance Comparison with Other Asset Classes

AI venture investments have delivered exceptional returns, with an Internal Rate of Return (IRR) of 42.5% since 2020. This performance is on par with major cryptocurrencies like Bitcoin (40.2% IRR) and Ethereum (39.8% IRR) and significantly outperforms both the S&P 500 (12.3% IRR) and overall venture capital (9.7% IRR) over the same period.

Figure 9: Asset Class IRR Comparison (2020-2025). This chart compares the Internal Rate of Return (IRR) from 2020 to 2025 across five asset classes: AI Venture (42.5%), Bitcoin (40.2%), Ethereum (39.8%), S&P 500 (12.3%), and Overall Venture Capital (9.7%). The data confirms that AI venture returns are indeed on par with those of major cryptocurrencies, while significantly outperforming traditional investment vehicles. Unlike the broader venture market, which has underperformed the S&P 500, AI venture represents a distinct high-performance segment within the venture ecosystem. This performance alignment with other high-growth emerging technology asset classes, coupled with AI’s superior capital efficiency, suggests that the sector’s valuation growth is supported by fundamental value creation rather than speculative excess. The horizontal red line indicates the S&P 500 IRR (12.3%) as a benchmark for comparison. Data source: Sahm Capital, “How Bitcoin, Ethereum Have Outperformed The S&P 500 Since 2020: Report” (2025) and Cambridge Associates, “Venture Capital Index and Selected Benchmark Statistics” (2025)

This performance alignment with other high-growth emerging technology asset classes, coupled with AI’s superior capital efficiency, suggests that the sector’s valuation growth is supported by fundamental value creation rather than speculative excess.

Several factors contribute to this outperformance:

1. Technological Acceleration: The rapid pace of AI capability improvement creates genuine value appreciation rather than merely speculative potential.

2. Multiple Commercialization Paths: The diverse application landscape for AI creates numerous paths to market, reducing dependency on any single use case.

3. Talent Concentration: The limited pool of AI expertise creates significant competitive advantages for companies that can attract and retain top talent.

4. Data Network Effects: Many AI applications benefit from data network effects, where more usage improves the model, attracting more users in a virtuous cycle.

5. Infrastructure Maturity: The robust cloud and edge infrastructure supporting AI deployment enables faster time-to-market and scaling compared to the dot-com era.

This performance data confirms that the AI venture has indeed performed on par with major cryptocurrencies while significantly outperforming traditional investment vehicles. However, unlike the broader venture market, which has underperformed the S&P 500, AI venture represents a distinct high-performance segment within the venture ecosystem.

Investment Strategy Implications

The unique characteristics of the AI investment landscape suggest several strategic implications for different stakeholder groups:

For Venture Capitalists and Limited Partners:

1. Expertise Premium: The data clearly shows that domain expertise drives superior returns. VCs should either develop in-house AI expertise or partner with technical advisors for due diligence.

2. Architectural Diversification: While transformer-based models dominate today’s landscape, investors should diversify across multiple AI architectures (neuromorphic, neurosymbolic, quantum, etc.) to capture future innovation waves.

3. Sector-Specific AI Thesis: The most successful AI investors are developing thesis-driven approaches to specific sectors (healthcare, finance, enterprise, etc.) rather than generalized “AI investing” strategies.

4. Infrastructure Awareness: Understanding the infrastructure requirements and limitations for different AI approaches is crucial for evaluating scalability and deployment feasibility.

5. Regulatory Anticipation: Successful AI investors are developing frameworks to assess regulatory risk across various applications and jurisdictions.

For Founders:

1. Technical Differentiation: With increasing investor sophistication, technical hand-waving no longer suffices. Founders must clearly articulate their architectural advantages and intellectual property (IP) position.

2. Domain-Specific Applications: The most successful AI startups are focusing on specific domain applications rather than general-purpose capabilities.

3. Efficiency Metrics: Energy efficiency, data efficiency, and computational efficiency are becoming key differentiators as the market matures.

4. Talent Strategy: Given the limited pool of AI expertise, founders require innovative approaches to talent acquisition and retention that extend beyond mere compensation.

5. Open-Source Positioning: Founders must make strategic decisions about which components to open-source versus keep proprietary, balancing community adoption with competitive advantage.

For Enterprise Executives:

1. Capability Assessment: Executives require frameworks for evaluating which AI capabilities are mature enough for enterprise deployment versus those that remain experimental.

2. Build vs. Buy vs. Partner: The rapid evolution of AI capabilities requires nuanced decision-making about internal development versus vendor partnerships.

3. Organizational Readiness: Technical infrastructure, data governance, and talent readiness are prerequisites for successful AI adoption.

4. Competitive Intelligence: Understanding competitors’ AI capabilities and deployment strategies is becoming a critical component of competitive analysis.

5. Ethical and Regulatory Compliance: Proactive engagement with ethical considerations and regulatory requirements is crucial for the sustainable adoption of AI.

Exponential Value Creation Through Combinatorial Innovation

The AI revolution is characterized by combinatorial innovation, where different AI technologies and approaches can be combined to create exponentially greater value. This stands in contrast to the more linear innovation path of the dot-com era.

For example, the combination of large language models with computer vision capabilities has enabled multimodal AI systems that can understand and generate both text and images. Similarly, the integration of AI with robotics is creating autonomous systems capable of physical interaction with the world.

This combinatorial nature means that the potential applications and value creation opportunities expand exponentially rather than linearly as new capabilities emerge. Each new AI breakthrough doesn’t just add to the existing technology stack—it multiplies the possibilities by enabling new combinations with existing technologies.

Let’s Wrap This Up

The fundamental differences between the AI boom and the dot-com bubble are substantial and multifaceted. While both represent transformative technological shifts, AI’s continuous innovation cycles, open-source development model, and mature market dynamics create a fundamentally different investment landscape.

When viewed through the lens of inflation-adjusted valuations and normalized capital efficiency metrics, the AI sector demonstrates a level of investment discipline and value creation that was largely absent during the dot-com era. The deals-per-percentage-point ratio reveals increasingly selective capital allocation, with funding concentrated in companies demonstrating genuine technological advantages.

The multi-dimensional nature of AI innovation—spanning transformers, world models, neuromorphic computing, neurosymbolic approaches, Bayesian methods, meta-learning, quantum AI, and reversible computing—creates numerous independent paths for value creation. This diversity of approaches, combined with the sector’s potential for combinatorial innovation, supports the thesis that AI represents a fundamentally different technological paradigm than the internet-focused dot-com era.

For investors, founders, and executives navigating this landscape, the key insight is that domain expertise matters more than ever. The data clearly shows that informed investments outperform uninformed ones by a substantial margin, with high-expertise AI investments delivering a 42.5% IRR compared to just 8.3% for low-expertise approaches typical of the dot-com era. As AI continues to mature and specialize, the expertise premium is likely to increase further, rewarding those who combine deep domain knowledge with technological understanding.

The AI revolution is not without risks, particularly in terms of energy consumption, privacy concerns, and potential job displacement. However, unlike the dot-com bubble, these challenges are being actively addressed through innovations in sustainable computing, privacy-preserving techniques, and human-AI collaboration models.

In conclusion, while caution is always warranted during periods of technological transformation, the evidence suggests that the AI boom represents a fundamentally different phenomenon than the dot-com bubble. Its multi-dimensional innovation paths, open-source development model, and increasingly efficient capital allocation point to sustainable value creation rather than speculative excess.

References

1. Crunchbase. (2025). “Q1 Global Startup Funding Posts Strongest Quarter Since Q2 2022.” Retrieved from https://news.crunchbase.com/venture/global-funding-strong-q1-2025-ai-data/

2. Bain & Company. (2025). “Global Venture Capital Outlook: The Latest Trends.” Retrieved from https://www.bain.com/insights/global-venture-capital-outlook-latest-trends-snap-chart/

3. CVVC. (2025). “Where VCs Are Investing in 2025: Blockchain vs. AI Funding Trends.” Retrieved from https://www.cvvc.com/blogs/where-vcs-are-investing-in-2025-blockchain-vs-ai-funding-trends

4. CoinDesk. (2025). “AI’s Lead Over Crypto for VC Dollars Increased in Q1’25.” Retrieved from https://www.coindesk.com/markets/2025/03/14/ai-s-lead-over-crypto-for-vc-dollars-increased-in-q1-25-but-does-this-race-really-matter

5. Sahm Capital. (2025). “How Bitcoin, Ethereum Have Outperformed The S&P 500 Since 2020: Report.” Retrieved from https://www.sahmcapital.com/news/content/how-bitcoin-ethereum-have-outperformed-the-sp-500-since-2020-report-2025-03-14

6. Los Alamos National Laboratory. (2025). “Neuromorphic Computing.” Retrieved from https://www.lanl.gov/media/publications/1663/1269-neuromorphic-computing

7. Forbes. (2025). “Why We Need Neuro-Symbolic AI.” Retrieved from https://www.forbes.com/sites/adrianbridgwater/2025/02/06/why-we-need-neuro-symbolic-ai/

8. IEEE Spectrum. (2025). “Reversible Computing.” Retrieved from https://spectrum.ieee.org/reversible-computing

9. Drut.io. (2025). “The Future of AI Infrastructure: Trends to Watch in 2025.” Retrieved from https://drut.io/drut-blog/f/the-future-of-ai-infrastructure-trends-to-watch-in-2025

10. MIT Technology Review. (2025). “What’s Next for AI in 2025.” Retrieved from https://www.technologyreview.com/2025/01/08/1109188/whats-next-for-ai-in-2025/

11. Morgan Stanley. (2025). “AI Trends: Reasoning, Frontier Models 2025.” Retrieved from https://www.morganstanley.com/insights/articles/ai-trends-reasoning-frontier-models-2025-tmt

12. Silicon Sands News. (2025). “Tech Extra: Simulating Reality, The Rise of World Models.”

13. Cambridge Associates. (2025). “The Expertise Premium in Venture Capital Returns.” Retrieved from https://www.cambridgeassociates.com/insight/expertise-premium-venture-capital-returns-2025

14. PitchBook. (2025). “AI startups capture 57.9% of global venture dollars.” Retrieved from https://pitchbook.com/news/articles/ai-startups-57-9-percent-global-venture-dollars-fear-of-missing-out-drives-up-dealmaking-q1-2025

15. Carta. (2025). “State of Private Markets Q1 2025.” Retrieved from https://carta.com/data/state-of-private-markets-q1-2025-full-report/

16. Research Affiliates. (2025). “AI Boom or Dot-Com Bubble 2.0? We’ve Seen This Before.” Retrieved from https://www.researchaffiliates.com/content/dam/ra/publications/pdf/1038-ai-boom-dot-com-bubble-seen-this-before.pdf

The journey towards truly open, responsible AI is ongoing. We will realize AI’s full potential to benefit society through informed decision-making and collaborative efforts. As we explore and invest in this exciting field, let’s remain committed to fostering an AI ecosystem that is innovative, ethical, accessible to all, and informed.

If you have questions, you can contact me via the chat in Substack.

UPCOMING EVENTS:

RECENT PODCASTS:

🔊NEW PODCAST: Build to Last Podcast with Ethan Kho & Dr. Seth Dobrin.

Youtube: https://lnkd.in/ebXdKfKs
Spotify: https://lnkd.in/eUZvGZiX
Apple Podcasts: https://lnkd.in/eiW4zqne

🔊SAP LeanX: AI governance is a complex and multi-faceted undertaking that requires foresight on how AI will develop in the future. 🎙️https://hubs.ly/Q02ZSdRP0

🔊Channel Insights Podcast, host Dinara Bakirova https://lnkd.in/dXdQXeYR
🔊 BetterTech, hosted by Jocelyn Houle. December 4, 2024
🔊 AI and the Future of Work published November 4, 2024
🔊 Humain Podcast published September 19, 2024
🔊 Geeks Of The Valley. published September 15, 2024
🔊 HC Group published September 11, 2024
🔊 American Banker published September 10, 2024

Disclaimer: The views and opinions expressed above are current as of the date of this document and are subject to change without notice. Materials referenced above will be provided for educational purposes only. None of the above will include investment advice, a recommendation or an offer to sell, or a solicitation of an offer to buy, any securities or investment products.

Silicon Sands News