Corporate Performance Management

Solution Notes

Lecture Notes
Author

Andy Weeger

Published

Aug 8, 2025

Modified

Dec 8, 2025

Financial analysis

Ratios

OpenAI

Profitability Ratios:

  • Gross Margin = (3,700 - 2,100) / 3,700 = 43.2%
  • Operating Margin = 200 / 3,700 = 5.4%
  • Net Margin = 150 / 3,700 = 4.1%
  • ROA = 150 / 5,000 = 3.0%
  • ROE = 150 / 2,500 = 6.0%

Efficiency & Liquidity Ratios:

  • Asset Turnover = 3,700 / 5,000 = 0.74
  • Current Ratio = 1,200 / 800 = 1.50
  • Quick Ratio = (1,200 - 150) / 800 = 1.31
  • Inventory Turnover = 2,100 / 150 = 14.0 times

Anthropic

Profitability Ratios:

  • Gross Margin = (1,000 - 650) / 1,000 = 35.0%
  • Operating Margin = 50 / 1,000 = 5.0%
  • Net Margin = 40 / 1,000 = 4.0%
  • ROA = 40 / 2,800 = 1.4%
  • ROE = 40 / 1,800 = 2.2%

Efficiency & Liquidity Ratios:

  • Asset Turnover = 1,000 / 2,800 = 0.36
  • Current Ratio = 900 / 400 = 2.25
  • Quick Ratio = (900 - 100) / 400 = 2.00
  • Inventory Turnover = 650 / 100 = 6.5 times

Discussion questions

Profitability

Which company is more profitable in percentage terms? What might explain the differences given both operate in AI?

OpenAI is more profitable across most metrics with higher Gross Margin (43.2% vs 35.0%), ROA (3.0% vs 1.4%), and ROE (6.0% vs 2.2%). Operating and Net Margins are similar (both around 4-5%).

Possible explanations:

  • Scale advantages: OpenAI’s higher revenue enables better economies of scale in computing
  • Revenue mix: ChatGPT consumer subscriptions may have higher margins than enterprise focus
  • Infrastructure efficiency: Microsoft partnership may provide advantageous computing costs
  • Product maturity: Established products command premium pricing
  • Development stage: Anthropic may be investing more heavily in R&D relative to revenue

Asset efficiency

Which company uses its assets more efficiently? What does this suggest about their business models or stage of development?

OpenAI uses assets more efficiently with Asset Turnover of 0.74 vs Anthropic’s 0.36, generating nearly twice as much revenue per euro of assets.

What this suggests:

OpenAI:

  • More mature business model with established revenue streams
  • Better capacity utilization
  • Closer to optimal operational scale

Anthropic:

  • Earlier development stage with assets not yet fully productive
  • May have raised significant capital ahead of revenue growth
  • Possible overcapacity built for anticipated future demand
  • Heavy investment in foundational research

OpenAI’s consumer-focused approach may generate faster asset returns, while Anthropic’s enterprise focus may require longer payback periods.

Inventory

Both companies hold GPU inventory. Which manages this more efficiently? What business factors might influence how much hardware inventory an AI company should maintain?

OpenAI manages GPU inventory more efficiently with Inventory Turnover of 14.0 times vs 6.5 times, cycling through inventory twice as fast.

Business factors influencing GPU inventory:

Operational:

  • Computational demand patterns (predictable vs. variable)
  • Training vs. inference balance
  • Cloud vs. owned infrastructure reliance

Strategic:

  • GPU availability and supply chain lead times
  • Technology refresh cycles and obsolescence risk
  • Scaling plans and growth expectations

Financial:

  • Capital availability for bulk purchasing
  • Opportunity cost of tied-up capital
  • Depreciation concerns as new generations emerge

Anthropic’s lower turnover might indicate building capacity ahead of demand or more conservative planning.

ROE

If you invested €1 million in equity in each company, which would generate better returns based on ROE? What cautions should you consider?

OpenAI would generate better returns with ROE of 6.0% vs 2.2% (€60,000 vs €22,000 annual return on €1M).

Critical cautions:

Leverage and sustainability:

  • ROE can be inflated by high debt levels
  • Historical returns may not predict future performance
  • Current margins may not be sustainable as competition intensifies

Growth stage:

  • Early-stage companies often suppress ROE through reinvestment
  • Anthropic’s lower ROE may reflect strategic growth prioritization
  • Private valuations may be many multiples of book equity

Risk factors:

  • ROE doesn’t account for risk differences
  • Significant intangible assets not captured in book value
  • Should consider volatility, competitive positioning, and execution risk

Better analysis would include multi-year trends, market valuations, qualitative factors, and forward projections.

Short-term obligations

Which company appears better positioned to meet short-term obligations? Why is this important for AI companies?

Anthropic is better positioned with Current Ratio of 2.25 vs 1.50 and Quick Ratio of 2.00 vs 1.31. Both are healthy, but Anthropic has a stronger cushion.

Why liquidity is critical for AI companies:

Operational demands:

  • High fixed computing costs (cloud bills, infrastructure) due monthly regardless of revenue
  • Cannot easily reduce costs without disrupting operations
  • Revenue may be unpredictable in rapidly evolving market

Growth and competition:

  • Cash burn during scaling (model training costs millions)
  • Cannot afford service interruptions that drive customers to competitors
  • Must continue investing to stay competitive

Financial flexibility:

  • Private companies can’t quickly access public markets
  • Raising new funding rounds takes months
  • Cloud providers may require significant deposits

OpenAI’s 1.50 ratio is adequate but tighter. Anthropic’s 2.25 provides significant safety margin, possibly reflecting recent funding or conservative management.

R&D

AI companies typically have high R&D costs not fully captured in COGS. How might this affect the profitability ratios?

R&D treatment: R&D costs appear as operating expenses (below gross profit), not in COGS.

In COGS: Direct computing for serving models, data acquisition
In operating expenses: Research scientist salaries, experimental training, safety research

Impact on ratios:

Gross Margin appears artificially high:

  • Both show healthy margins (43.2%, 35.0%)
  • Doesn’t reflect true cost of developing AI capabilities
  • Makes AI companies look like software (high margins) rather than biotech (high R&D)

Operating Margin better reflects reality:

  • Much lower (5.4%, 5.0%), showing true profitability challenge
  • Includes R&D expenses
  • Better metric for AI company comparison

ROA implications:

  • R&D creates intangible assets (models) not capitalized
  • Assets are understated, making returns appear lower
  • Heavy R&D spending today creates competitive moats tomorrow

Better approach: Focus on operating margin as primary metric, assess R&D productivity, and understand that current expenses create future advantages.

Scale

These companies are scaling rapidly. How might growth stage affect the interpretation of ROA and Asset Turnover?

Growth stage distortions:

Timing mismatches:

  • Assets deployed before revenue materializes
  • Infrastructure and teams built months before revenue impact
  • Creates temporarily depressed metrics during scaling
  • “J-curve” effect: metrics decline initially, then improve rapidly

Capacity building:

  • OpenAI’s higher Asset Turnover (0.74): More mature, operating near capacity
  • Anthropic’s lower turnover (0.36): Building capacity for 2-3x growth, not at steady-state

Different asset bases:

  • Recent funding inflates assets relative to operations
  • Cash sits as “unproductive” until deployed
  • Makes metrics look worse for recently-funded companies

Better analytical questions:

Instead of “Is ROA high enough?”:

  • Is ROA improving quarter-over-quarter?
  • What’s the trajectory as infrastructure comes online?
  • Are assets being deployed productively?

Instead of “Is Asset Turnover strong?”:

  • What’s capacity utilization rate?
  • How much growth runway do current assets support?
  • What’s expected turnover at maturity?

Investment perspective: Don’t penalize low current metrics if intentional for growth. Look for operating leverage and compare to peers at similar stages.

Revenue

OpenAI has partnership revenue from Microsoft, while Anthropic is developing enterprise solutions. How might different business models influence these financial ratios?

Business model characteristics:

OpenAI:

  • Consumer subscriptions (high volume, lower per-unit)
  • API business (usage-based, scalable)
  • Microsoft partnership (infrastructure advantages)

Anthropic:

  • Enterprise solutions (lower volume, higher per-unit, longer sales cycles)
  • API business (similar to OpenAI)
  • Safety focus (appeals to regulated industries)

Impact on ratios:

Gross Margin (43.2% vs 35.0%):

  • OpenAI’s economies of scale and consumer mix drive higher margins
  • Anthropic’s enterprise support and customization increase COGS
  • Enterprise deals require dedicated teams and implementations

Asset Turnover (0.74 vs 0.36):

  • Self-service products generate revenue faster per asset
  • Enterprise sales have 3-12 month cycles; infrastructure built before contracts close
  • Expensive sales teams are assets not generating immediate returns

Liquidity (Anthropic stronger at 2.25 vs 1.50):

  • Enterprise contracts often have milestone payments creating cash reserves
  • Fewer, larger customers create lumpier but larger cash inflows
  • May reflect capital raised for longer sales cycles

Strategic trade-offs:

OpenAI advantages: Faster growth, better asset utilization, viral adoption, lower CAC
OpenAI risks: Higher churn, pricing pressure, commoditization

Anthropic advantages: Stickier customers, higher lifetime value, premium pricing
Anthropic risks: Slower growth, higher CAC, execution complexity

Intangible assets

Both companies likely have significant intangible assets (models, datasets, talent) not fully reflected on balance sheets. How does this limitation affect your analysis?

Intangible assets not captured:

Trained AI models:

  • Development costs billions but expensed as R&D
  • Not capitalized on balance sheet
  • True “asset” value far exceeds book value

Proprietary datasets:

  • Curated training data is competitive advantage
  • Data partnerships and licenses not reflected
  • Cleaning and curation efforts expensed immediately

Human capital:

  • Top AI researchers worth millions in competitive value
  • Team expertise and culture not quantified
  • Retention risk not visible in financials

Brand and market position:

  • “ChatGPT” brand recognition has enormous value
  • First-mover advantages not captured
  • Reputation in safety/ethics matters for enterprise sales

How this affects analysis:

Asset-based ratios (ROA, Asset Turnover) are misleading:

  • Denominators (Total Assets) dramatically understated
  • True ROA likely much higher than calculated
  • Companies with better models show artificially low ratios

Equity value misrepresented:

  • Book equity far below market value
  • ROE doesn’t reflect true return on capital invested
  • Private valuations (billions) vs. book equity mismatch

Competitive position unclear:

  • Can’t assess “moat” strength from financials
  • Quality differences between models not quantified
  • Strategic advantages invisible

Better analytical approach:

Supplement with qualitative assessment:

  • Model benchmark performance (MMLU, HumanEval scores)
  • Research output (papers, citations, breakthroughs)
  • Team quality (notable hires, retention rates)
  • Market position indicators (API adoption, customer logos)

Use market-based metrics:

  • Funding round valuations as proxy for true value
  • Revenue multiples from comparable companies
  • Cost per model capability improvement

Focus on operating metrics:

  • Model performance per dollar spent
  • Revenue per researcher/engineer
  • Customer acquisition and retention rates
  • Time-to-market for new capabilities

Investment implication: Traditional financial ratios significantly understate the value and competitive position of AI companies. Must combine quantitative financials with qualitative assessment of intangible assets and capabilities.