Chat with DeepSeek about Impacts on US Venture Capital
A conversation with an AI about AI
On 23 January 2025 I had two ‘conversations’ about AI with two AI platforms: one with ChatGPT and one with DeepSeek. Again, the questions related to the impact of the launch of DeepSeek on the business models and valuation assumptions of existing proprietary alternatives. The upshot was an estimated impairment in the range of 20-50%. The ‘conversation’ with DeepSeek is reproduced below verbatim.
WP: With the release of DeepSeek as an open source platform, and with its pricing at rates substantially lower than comparable pricing models from US models and corporations, it seems likely or inevitable that the entire valuations foundation for American proprietary AI or large language model systems and corporations will need to be reviewed. In this context, the valuations that were applied to the principal American AI projects in the past seem to be no longer valid, resulting in the likelihood or indeed inevitability of valuation impairment. Is this the case? If so, what are the likely implications for American venture capital and private equity exposed to the sector in the US?
DS: The release of DeepSeek as an open-source platform, coupled with its significantly lower pricing compared to proprietary US-based AI models, could indeed disrupt the valuation foundations of American AI corporations and large language model (LLM) systems. Here's an analysis of the situation and its potential implications:
1. Valuation Impairment for US AI Projects
Competitive Pressure: DeepSeek’s open-source nature and lower pricing create competitive pressure on proprietary US AI models. If DeepSeek or similar platforms offer comparable or superior performance at a fraction of the cost, the perceived value of proprietary systems may decline.
Revenue Model Challenges: Many US AI companies rely on subscription-based or usage-based pricing models. If open-source alternatives like DeepSeek gain traction, these revenue streams could shrink, leading to downward revisions in revenue projections and, consequently, valuations.
Market Share Erosion: Open-source platforms often attract a broad user base, including developers, startups, and enterprises, due to their cost-effectiveness and flexibility. This could erode the market share of proprietary US AI systems, further impacting their valuations.
2. Implications for Venture Capital (VC) and Private Equity (PE)
Downward Valuation Adjustments: VC and PE firms with significant exposure to US AI companies may face downward adjustments in the valuations of their portfolio companies. This could lead to markdowns in the value of their investments.
Increased Scrutiny on New Investments: Investors may become more cautious about funding new AI ventures, particularly those relying on proprietary models with high development and operational costs. They may shift focus to startups leveraging open-source technologies or those with innovative business models.
Portfolio Restructuring: VC and PE firms might reassess their portfolios, potentially divesting from companies heavily reliant on proprietary AI models and reallocating capital to firms that can integrate or compete with open-source solutions.
Focus on Differentiation: Investors may prioritize AI companies that offer unique value propositions beyond just the model itself, such as superior data pipelines, domain-specific applications, or robust ecosystem integrations.
3. Broader Market Implications
Acceleration of Open-Source Adoption: The success of DeepSeek could accelerate the adoption of open-source AI platforms across industries, further challenging the dominance of proprietary systems.
Innovation and Collaboration: The rise of open-source AI could foster greater innovation and collaboration, as developers and organizations build on shared technologies. This might lead to faster advancements in the field but could also dilute the competitive edge of proprietary players.
Regulatory and Ethical Considerations: As open-source AI gains prominence, regulators may focus more on ensuring ethical use, transparency, and accountability, which could impact both open-source and proprietary models.
4. Strategic Responses for US AI Companies
Cost Optimization: Proprietary AI companies may need to reduce costs to remain competitive, potentially through efficiency improvements or leveraging open-source components themselves.
Focus on Value-Added Services: Companies could differentiate by offering value-added services, such as customization, support, or integration with existing enterprise systems.
Partnerships and Ecosystem Building: Collaborating with open-source communities or integrating open-source models into their offerings could help proprietary players stay relevant.
Conclusion
The emergence of DeepSeek and similar open-source AI platforms is likely to disrupt the valuation landscape for US AI companies, particularly those reliant on proprietary models. Venture capital and private equity investors exposed to the sector may need to reassess their portfolios and investment strategies to navigate this shifting landscape. While challenges exist, the situation also presents opportunities for innovation, collaboration, and the development of new business models in the AI space.
WP: What level or degree of downward valuation adjustments can we reasonably expect given the dramatic pricing difference? Also take into account the operating costs of the different models (energy consumed for a given computational output for a comparable set of problems) and what this may mean for the relative valuations of American AI corporations.
DS: The degree of downward valuation adjustments for American AI corporations in response to the dramatic pricing differences introduced by open-source platforms like DeepSeek depends on several factors, including the competitive landscape, operating costs, and the ability of proprietary models to differentiate themselves. Here’s a detailed analysis:
1. Pricing Differences and Valuation Impact
DeepSeek's Pricing Advantage: If DeepSeek or similar open-source platforms offer comparable performance at a fraction of the cost, the pricing power of proprietary US AI models will be severely constrained. This could lead to significant revenue compression for companies relying on high-margin pricing models.
Valuation Multiples: AI companies are often valued based on revenue multiples, particularly in high-growth sectors. If revenue growth slows or margins shrink due to competitive pricing pressure, valuation multiples could contract. For example, if a company's revenue growth rate drops from 50% to 20% due to market share loss, its price-to-sales (P/S) multiple might decline from 20x to 10x or lower, depending on market sentiment.
Magnitude of Adjustments: In extreme cases, valuations could drop by 30-50% or more for companies that fail to differentiate themselves or adapt to the new competitive landscape. Companies with high operating costs and limited pricing flexibility would be most vulnerable.
2. Operating Costs and Energy Efficiency
Energy Consumption and Computational Efficiency: Open-source models like DeepSeek may have lower operating costs if they are more energy-efficient or optimized for cost-effective hardware. Proprietary models, which often rely on expensive infrastructure (e.g., custom GPUs or TPUs), could face higher operational expenses, further eroding margins.
Cost-to-Serve Comparison: If DeepSeek’s cost-to-serve (including energy and computational costs) is significantly lower, it could sustain lower pricing while maintaining profitability. Proprietary models with higher costs would struggle to compete, leading to margin compression and reduced profitability.
Implications for Valuations: Companies with higher operating costs would face greater valuation impairments, as investors factor in lower future cash flows. For example, if a company's operating margins shrink from 30% to 10% due to competitive and cost pressures, its earnings-based valuation (e.g., P/E ratio) could decline proportionally.
3. Competitive Differentiation and Mitigating Factors
Unique Value Propositions: Proprietary AI companies that can differentiate themselves through superior performance, domain-specific expertise, or unique features may mitigate some of the downward pressure on valuations. For example, a company specializing in healthcare or legal AI applications might retain higher valuations if its models offer unmatched accuracy or regulatory compliance.
Ecosystem Lock-In: Companies with strong ecosystem integrations (e.g., Microsoft's OpenAI integration) may retain customer loyalty despite higher pricing, reducing the risk of valuation impairment.
Innovation and R&D: Continued investment in R&D to improve model efficiency, reduce costs, and enhance capabilities could help proprietary players maintain their competitive edge.
4. Sector-Wide Implications
Early-Stage Startups: Early-stage AI startups with high burn rates and unproven business models may face the most severe valuation adjustments, as investors become more risk-averse.
Established Players: Larger, established AI companies (e.g., OpenAI, Anthropic) may experience moderate valuation declines but could leverage their brand, customer base, and resources to adapt.
Investor Sentiment: A broader reassessment of AI valuations could lead to a sector-wide repricing, particularly if open-source platforms gain significant market share. This could result in a 20-40% sector-wide valuation decline in the short to medium term.
5. Long-Term Outlook
Consolidation: The sector may see consolidation as weaker players struggle to compete, leading to mergers or acquisitions by stronger firms.
Shift to Hybrid Models: Proprietary AI companies might adopt hybrid models, incorporating open-source components while offering premium services to maintain profitability.
Increased Focus on Efficiency: Companies will likely prioritize energy efficiency and cost optimization to remain competitive, which could improve long-term sustainability.
Quantitative Estimate of Valuation Adjustments
Based on the above factors, here’s a rough estimate of potential valuation adjustments:
High-Cost, Undifferentiated Players: 40-60% decline in valuations due to revenue compression, margin erosion, and loss of competitive advantage.
Differentiated Players with Strong Ecosystems: 10-30% decline, as they retain some pricing power and customer loyalty.
Early-Stage Startups: 50-70% decline or more, as investors reassess the viability of their business models in a more competitive landscape.
Conclusion
The dramatic pricing difference introduced by open-source platforms like DeepSeek, combined with operating cost disparities, is likely to lead to significant downward valuation adjustments for many American AI corporations. The extent of these adjustments will depend on each company's ability to differentiate itself, optimize costs, and adapt to the new competitive reality. Venture capital and private equity investors should brace for a sector-wide repricing and consider reallocating capital to companies with sustainable competitive advantages or those leveraging open-source innovations effectively.
WP: What is the pricing differential between DeepSeek and its comparable competitive offerings?
DS: As of my knowledge cutoff in October 2023, specific pricing details for DeepSeek and its direct competitors (e.g., OpenAI’s GPT-4, Anthropic's Claude, Google’s Bard, etc.) are not publicly available in granular form. However, I can provide a general analysis of the pricing differentials based on trends in the AI industry and the typical cost structures of open-source versus proprietary models.
1. Open-Source vs. Proprietary Pricing Models
Open-Source Platforms (e.g., DeepSeek):
Open-source models often have little to no licensing costs, as they are freely available for use and modification.
Users typically pay only for infrastructure costs (e.g., cloud computing, energy, and hardware) when deploying these models.
Some open-source platforms may offer premium support or enterprise features at a fraction of the cost of proprietary solutions.
Estimated Cost: For a comparable computational output, open-source models might cost 10-30% of proprietary solutions, depending on optimization and scale.
Proprietary Models (e.g., OpenAI, Anthropic, Google):
Proprietary models often charge based on usage (e.g., per token or API call) or subscription plans.
Pricing is typically higher to recoup R&D costs, infrastructure investments, and generate profits.
For example, OpenAI's GPT-4 API costs approximately $0.03–$0.06 per 1,000 tokens for input and $0.06–$0.12 per 1,000 tokens for output, depending on the model version and usage tier.
Estimated Cost: Proprietary models can be 3-10x more expensive than open-source alternatives for equivalent tasks.
2. DeepSeek’s Pricing Advantage
If DeepSeek follows the typical open-source model, its pricing could be significantly lower than proprietary offerings. For example:
Infrastructure Costs Only* Users deploying DeepSeek would primarily pay for cloud or on-premises infrastructure, which could be optimized for cost efficiency.
No Licensing Fees: Unlike proprietary models, DeepSeek would not charge for access to the model itself.
Enterprise Features: If DeepSeek offers premium features (e.g., enhanced support, fine-tuning, or integration tools), these might still be priced lower than proprietary alternatives.
3. Comparative Cost Analysis
Proprietary Model Example:
OpenAI GPT-4: ~$0.03–$0.12 per 1,000 tokens.
Anthropic Claude: ~$0.10–$0.20 per 1,000 tokens.
Google Bard: Pricing varies but is generally in line with other proprietary models.
Open-Source Model Example (DeepSeek):
Infrastructure costs: ~$0.01–$0.03 per 1,000 tokens (depending on optimization and scale).
No licensing fees: $0.
Pricing Differential:
DeepSeek could be 70-90% cheaper than proprietary models for equivalent tasks, assuming efficient infrastructure usage.
4. Energy and Computational Efficiency
Open-source models like DeepSeek may be more energy-efficient if they are optimized for cost-effective hardware or leverage community-driven improvements.
Proprietary models often rely on expensive, custom hardware (e.g., GPUs, TPUs) and may have higher energy consumption, further widening the cost gap.
5. Implications for Users
Cost Savings: Enterprises and developers using DeepSeek could achieve significant cost savings, especially at scale.
Flexibility: Open-source models allow for customization and fine-tuning, which can further reduce costs for specific use cases.
Adoption Incentives: The lower cost of open-source models could accelerate their adoption, particularly among startups, researchers, and cost-sensitive organizations.
Conclusion
The pricing differential between DeepSeek and proprietary offerings like OpenAI’s GPT-4 or Anthropic’s Claude is likely to be substantial, with DeepSeek potentially costing **70-90% less** for comparable tasks. This dramatic cost advantage, combined with the flexibility and customization options of open-source models, could disrupt the AI market and force proprietary players to rethink their pricing strategies and cost structures.
WP: What was the last valuation for OpenAI and other comparable corporations? What were the marketshare and pricing / revenue assumptions behind these valuations? What is the price elasticity of demand in the AI market? Assuming that for the vast majority of tasks, users will gravitate to the least cost option, the 70-90% price advantage of DeepSeek over comparable alternatives could undermine market share for these alternatives.
DS: The valuations of major AI corporations like OpenAI, Anthropic, and others are based on a combination of market share, revenue projections, and the perceived potential of their technologies. However, the emergence of open-source alternatives like DeepSeek, with a significant price advantage, could disrupt these valuations by undermining market share and revenue assumptions. Let’s break this down:
1. Valuations of Major AI Corporations
OpenAI:
Latest Valuation: As of 2023, OpenAI was valued at $29 billion following a funding round led by Microsoft and other investors.
Revenue Assumptions: OpenAI’s valuation was based on projections of rapid adoption of its API services (e.g., GPT-4), enterprise partnerships, and integration into products like Microsoft’s Azure and Office suite. Revenue was expected to grow significantly as AI adoption expanded across industries.
Market Share: OpenAI has been a leader in the generative AI space, with a significant share of the market for large language models (LLMs) and AI APIs.
Anthropic:
Latest Valuation: Anthropic was valued at $4.1 billion in its latest funding round.
Revenue Assumptions: Anthropic’s valuation was based on its focus on ethical AI and its Claude model, which targets enterprise customers willing to pay a premium for safer, more reliable AI systems.
Market Share: Anthropic has a smaller but growing share of the enterprise AI market, particularly in sectors like healthcare, legal, and finance.
Cohere:
Latest Valuation: Cohere was valued at $2.2 billion in 2023.
Revenue Assumptions: Cohere’s valuation was driven by its focus on enterprise-grade AI solutions and partnerships with major cloud providers.
Market Share: Cohere has been gaining traction in the enterprise market, particularly for custom AI solutions.
2. Market Share and Pricing Assumptions
Market Share:
Proprietary AI companies like OpenAI and Anthropic have relied on their first-mover advantage, brand recognition, and enterprise partnerships to capture significant market share.
However, their market share is vulnerable to disruption from open-source alternatives like DeepSeek, which offer comparable performance at a fraction of the cost.
Pricing and Revenue Assumptions:
- Proprietary models are priced at a premium to recoup R&D costs and generate profits. For example:
- OpenAI’s GPT-4 API costs $0.03–$0.12 per 1,000 tokens.
- Anthropic’s Claude API costs $0.10–$0.20 per 1,000 tokens.
Revenue projections assume continued growth in API usage, enterprise adoption, and expansion into new markets.
3. Price Elasticity of Demand in the AI Market
Definition: Price elasticity of demand measures how sensitive the quantity demanded is to changes in price. In the AI market, demand is likely to be highly elastic for many use cases, as users can easily switch to lower-cost alternatives.
Factors Influencing Elasticity:
Task Sensitivity: For tasks where performance differences between models are minimal, users are likely to gravitate toward the least-cost option (e.g., DeepSeek).
Enterprise Needs: Enterprises may be less price-sensitive for mission-critical applications requiring high reliability, security, or compliance, but even they will seek cost-effective solutions.
Developer Preferences: Developers and startups, who are often cost-sensitive, are likely to adopt open-source models like DeepSeek to reduce expenses.
4. Impact of DeepSeek’s Price Advantage
Market Share Erosion:
DeepSeek’s 70-90% price advantage could lead to significant market share erosion for proprietary models, particularly in cost-sensitive segments like startups, academia, and small-to-medium enterprises (SMEs).
Even enterprise customers may pressure proprietary vendors to lower prices or risk losing business to open-source alternatives.
Revenue Compression:
If DeepSeek captures a substantial portion of the market, proprietary AI companies could face revenue compression, as their high-margin pricing models become unsustainable.
For example, if OpenAI loses 30% of its market share to DeepSeek, its revenue growth projections would need to be revised downward, leading to valuation impairments.
Valuation Adjustments:
Valuations for proprietary AI companies could decline by 20-50% or more, depending on the extent of market share loss and revenue compression.
Early-stage startups with unproven business models may face even steeper valuation declines.
5. Strategic Implications
Proprietary AI Companies:
May need to lower prices to remain competitive, which could erode margins and profitability.
Could focus on differentiating through superior performance, domain-specific expertise, or value-added services (e.g., support, compliance, integration).
Open-Source Platforms (e.g., DeepSeek):
Likely to gain rapid adoption, particularly among cost-sensitive users and developers.
Could monetize through premium support, enterprise features, or ecosystem partnerships.
Investors:
- May shift capital toward open-source AI projects or companies that can integrate open-source models effectively.
- Will likely demand greater scrutiny of revenue projections and competitive positioning for proprietary AI investments.
Conclusion
The 70-90% price advantage of DeepSeek over proprietary alternatives like OpenAI and Anthropic could significantly undermine the market share and revenue assumptions behind their valuations. Given the high price elasticity of demand in the AI market, users are likely to gravitate toward the least-cost option for the vast majority of tasks. This could lead to valuation impairments of 20-50% or more for proprietary AI companies, particularly if they fail to differentiate themselves or adapt to the new competitive landscape. Investors and stakeholders should closely monitor these dynamics and adjust their strategies accordingly.