Quantifying the AI Bubble Final Project

Are AI Company Valuations Growing Faster Than Financial Performance?

library(tidyverse)

── Attaching core tidyverse packages ──────────────────────── tidyverse 2.0.0 ──
✔ dplyr     1.2.1     ✔ readr     2.1.6
✔ forcats   1.0.1     ✔ stringr   1.6.0
✔ ggplot2   4.0.3     ✔ tibble    3.3.1
✔ lubridate 1.9.4     ✔ tidyr     1.3.2
✔ purrr     1.2.2     
── Conflicts ────────────────────────────────────────── tidyverse_conflicts() ──
✖ dplyr::filter() masks stats::filter()
✖ dplyr::lag()    masks stats::lag()
ℹ Use the conflicted package (<http://conflicted.r-lib.org/>) to force all conflicts to become errors

library(tidyquant)

Registered S3 method overwritten by 'quantmod':
  method            from
  as.zoo.data.frame zoo 
── Attaching core tidyquant packages ─────────────────────── tidyquant 1.0.12 ──
✔ PerformanceAnalytics 2.1.0      ✔ TTR                  0.24.4
✔ quantmod             0.4.28     ✔ xts                  0.14.1── Conflicts ────────────────────────────────────────── tidyquant_conflicts() ──
✖ zoo::as.Date()                 masks base::as.Date()
✖ zoo::as.Date.numeric()         masks base::as.Date.numeric()
✖ dplyr::filter()                masks stats::filter()
✖ xts::first()                   masks dplyr::first()
✖ dplyr::lag()                   masks stats::lag()
✖ xts::last()                    masks dplyr::last()
✖ PerformanceAnalytics::legend() masks graphics::legend()
✖ quantmod::summary()            masks base::summary()
ℹ Use the conflicted package (<http://conflicted.r-lib.org/>) to force all conflicts to become errors

library(lubridate)
library(scales)

Attaching package: 'scales'

The following object is masked from 'package:purrr':

    discard

The following object is masked from 'package:readr':

    col_factor

library(gt)
library(broom)

Project Overview

This project examines whether stock market valuations of major AI-focused technology companies have grown faster than their underlying financial performance. With increased investor excitement around artificial intelligence, companies such as NVIDIA, Microsoft, Alphabet, Amazon, and Meta have experienced major market growth. This analysis compares stock price growth against business fundamentals such as revenue and net income to evaluate whether market performance appears aligned with financial performance.

The project follows the OSEMN data science workflow:

1. Obtain financial market and company fundamentals data
   
2. Scrub and clean the data into comparable formats
   
3. Explore trends in stock prices, revenue, and net income
   
4. Model the relationship between financial performance and stock growth
   
5. Interpret whether valuation growth appears supported by fundamentals

First let’s define the AI-related companies for this project

ai_tickers <- c("NVDA", "MSFT", "GOOGL", "AMZN", "META")
stock_prices <- tq_get(
  ai_tickers,
  get = "stock.prices",
  from = "2020-01-01",
  to = Sys.Date()
)
# Preview the data
glimpse(stock_prices)

Rows: 7,980
Columns: 8
$ symbol   <chr> "NVDA", "NVDA", "NVDA", "NVDA", "NVDA", "NVDA", "NVDA", "NVDA…
$ date     <date> 2020-01-02, 2020-01-03, 2020-01-06, 2020-01-07, 2020-01-08, …
$ open     <dbl> 5.96875, 5.87750, 5.80800, 5.95500, 5.99400, 6.09625, 6.18325…
$ high     <dbl> 5.99775, 5.94575, 5.93175, 6.04425, 6.05100, 6.14825, 6.21375…
$ low      <dbl> 5.91800, 5.85250, 5.78175, 5.90975, 5.95375, 6.02150, 6.09375…
$ close    <dbl> 5.99775, 5.90175, 5.92650, 5.99825, 6.00950, 6.07550, 6.10800…
$ volume   <dbl> 237536000, 205384000, 262636000, 314856000, 277108000, 255112…
$ adjusted <dbl> 5.970753, 5.875186, 5.899825, 5.971251, 5.982451, 6.048154, 6…

Now let’s pull company financial fundamentals. This gives us revenue + net income directly from financial statements.

library(httr2)
library(jsonlite)

Attaching package: 'jsonlite'

The following object is masked from 'package:purrr':

    flatten

company_map <- tibble(
  symbol = c("NVDA", "MSFT", "GOOGL", "AMZN", "META"),
  company = c("NVIDIA", "Microsoft", "Alphabet", "Amazon", "Meta"),
  cik = c("0001045810", "0000789019", "0001652044", "0001018724", "0001326801")
)

get_sec_revenue <- function(symbol, company, cik) {
  
  url <- paste0(
    "https://data.sec.gov/api/xbrl/companyfacts/CIK",
    cik,
    ".json"
  )
  
  response <- request(url) %>%
    req_user_agent("MSDS 607 project analysis") %>%
    req_perform()
  
  data <- response %>%
    resp_body_json()
  
  revenue_options <- c(
    "Revenues",
    "RevenueFromContractWithCustomerExcludingAssessedTax",
    "SalesRevenueNet",
    "SalesRevenueGoodsNet",
    "SalesRevenueServicesNet"
  )
  
  all_revenue_tables <- list()
  
  for (field in revenue_options) {
    
    if (field %in% names(data$facts$`us-gaap`)) {
      
      possible_units <- data$facts$`us-gaap`[[field]]$units
      
      if ("USD" %in% names(possible_units)) {
        
        temp_tbl <- bind_rows(possible_units$USD)
        
        if ("form" %in% names(temp_tbl)) {
          
          cleaned_tbl <- temp_tbl %>%
            filter(form == "10-K") %>%
            mutate(
              start = as.Date(start),
              end = as.Date(end),
              filed = as.Date(filed),
              duration_days = as.numeric(end - start)
            ) %>%
            filter(duration_days >= 300, duration_days <= 370) %>%
            filter(year(end) == fy) %>%
            select(fy, filed, val) %>%
            mutate(
              symbol = symbol,
              company = company,
              revenue = val,
              revenue_field = field
            ) %>%
            select(symbol, company, fy, filed, revenue, revenue_field)
          
          all_revenue_tables[[field]] <- cleaned_tbl
        }
      }
    }
  }
  
  best_tbl <- bind_rows(all_revenue_tables) %>%
    group_by(revenue_field) %>%
    mutate(field_row_count = n()) %>%
    ungroup() %>%
    arrange(desc(field_row_count)) %>%
    group_by(symbol, company, fy) %>%
    slice(1) %>%
    ungroup()
  
  return(best_tbl)
}

revenue_data <- pmap_dfr(
  company_map,
  get_sec_revenue
)

revenue_clean <- revenue_data %>%
  group_by(symbol, company, fy) %>%
  arrange(desc(filed)) %>%
  slice(1) %>%
  ungroup() %>%
  filter(fy >= 2020) %>%
  arrange(symbol, fy)

revenue_clean

# A tibble: 31 × 7
   symbol company     fy filed           revenue revenue_field   field_row_count
   <chr>  <chr>    <int> <date>            <dbl> <chr>                     <int>
 1 AMZN   Amazon    2020 2021-02-03 386064000000 RevenueFromCon…               8
 2 AMZN   Amazon    2021 2022-02-04 469822000000 RevenueFromCon…               8
 3 AMZN   Amazon    2022 2023-02-03 513983000000 RevenueFromCon…               8
 4 AMZN   Amazon    2023 2024-02-02 574785000000 RevenueFromCon…               8
 5 AMZN   Amazon    2024 2025-02-07 637959000000 RevenueFromCon…               8
 6 AMZN   Amazon    2025 2026-02-06 716924000000 RevenueFromCon…               8
 7 GOOGL  Alphabet  2020 2021-02-03 182527000000 Revenues                      6
 8 GOOGL  Alphabet  2021 2022-02-02 257637000000 Revenues                      6
 9 GOOGL  Alphabet  2022 2023-02-03 282836000000 RevenueFromCon…               5
10 GOOGL  Alphabet  2023 2024-01-31 307394000000 RevenueFromCon…               5
# ℹ 21 more rows

It looks like this has duplicate SEC records per fiscal year, so before we pull net income, let’s clean revenue first.

revenue_clean <- revenue_data %>%
  mutate(
    filed = as.Date(filed)
  ) %>%
  group_by(symbol, company, fy) %>%
  arrange(desc(filed), desc(revenue)) %>%
  slice(1) %>%
  ungroup() %>%
  filter(fy >= 2020) %>%
  arrange(symbol, fy)

revenue_clean

# A tibble: 31 × 7
   symbol company     fy filed           revenue revenue_field   field_row_count
   <chr>  <chr>    <int> <date>            <dbl> <chr>                     <int>
 1 AMZN   Amazon    2020 2021-02-03 386064000000 RevenueFromCon…               8
 2 AMZN   Amazon    2021 2022-02-04 469822000000 RevenueFromCon…               8
 3 AMZN   Amazon    2022 2023-02-03 513983000000 RevenueFromCon…               8
 4 AMZN   Amazon    2023 2024-02-02 574785000000 RevenueFromCon…               8
 5 AMZN   Amazon    2024 2025-02-07 637959000000 RevenueFromCon…               8
 6 AMZN   Amazon    2025 2026-02-06 716924000000 RevenueFromCon…               8
 7 GOOGL  Alphabet  2020 2021-02-03 182527000000 Revenues                      6
 8 GOOGL  Alphabet  2021 2022-02-02 257637000000 Revenues                      6
 9 GOOGL  Alphabet  2022 2023-02-03 282836000000 RevenueFromCon…               5
10 GOOGL  Alphabet  2023 2024-01-31 307394000000 RevenueFromCon…               5
# ℹ 21 more rows

Some companies/years are missing, especially Alphabet and NVIDIA. Let’s fix that.

annual_stock_growth <- stock_prices %>%
  mutate(year = year(date)) %>%
  group_by(symbol, year) %>%
  summarize(
    first_price = first(adjusted),
    last_price = last(adjusted),
    stock_growth_pct = (last_price / first_price - 1) * 100,
    .groups = "drop"
  ) %>%
  filter(year >= 2020) %>%
  arrange(symbol, year)

annual_stock_growth

# A tibble: 35 × 5
   symbol  year first_price last_price stock_growth_pct
   <chr>  <dbl>       <dbl>      <dbl>            <dbl>
 1 AMZN    2020        94.9      163.             71.6 
 2 AMZN    2021       159.       167.              4.64
 3 AMZN    2022       170.        84             -50.7 
 4 AMZN    2023        85.8      152.             77.0 
 5 AMZN    2024       150.       219.             46.3 
 6 AMZN    2025       220.       231.              4.81
 7 AMZN    2026       226.       273.             20.4 
 8 GOOGL   2020        67.9       86.9            28.1 
 9 GOOGL   2021        85.6      144.             67.8 
10 GOOGL   2022       144.        87.5           -39.1 
# ℹ 25 more rows

This is much better and cleaner. Let’s pull net income from SEC now for these companies.

get_sec_net_income <- function(symbol, company, cik) {
  
  url <- paste0(
    "https://data.sec.gov/api/xbrl/companyfacts/CIK",
    cik,
    ".json"
  )
  
  response <- request(url) %>%
    req_user_agent("MSDS 607 project analysis") %>%
    req_perform()
  
  data <- response %>%
    resp_body_json()
  
  net_income_data <- data$facts$`us-gaap`$NetIncomeLoss$units$USD
  
  net_income_tbl <- bind_rows(net_income_data) %>%
    filter(form == "10-K") %>%
    mutate(
      start = as.Date(start),
      end = as.Date(end),
      filed = as.Date(filed),
      duration_days = as.numeric(end - start)
    ) %>%
    filter(duration_days >= 300, duration_days <= 370) %>%
filter(year(end) == fy) %>%
select(fy, filed, val) %>%
    mutate(
      symbol = symbol,
      company = company,
      net_income = val
    ) %>%
    select(symbol, company, fy, filed, net_income)
  
  return(net_income_tbl)
}

net_income_data <- pmap_dfr(
  company_map,
  get_sec_net_income
)

net_income_clean <- net_income_data %>%
  group_by(symbol, company, fy) %>%
  arrange(desc(filed)) %>%
  slice(1) %>%
  ungroup() %>%
  filter(fy >= 2020) %>%
  arrange(symbol, fy)

net_income_clean

# A tibble: 31 × 5
   symbol company     fy filed       net_income
   <chr>  <chr>    <int> <date>           <dbl>
 1 AMZN   Amazon    2020 2021-02-03 21331000000
 2 AMZN   Amazon    2021 2022-02-04 33364000000
 3 AMZN   Amazon    2022 2023-02-03 -2722000000
 4 AMZN   Amazon    2023 2024-02-02 30425000000
 5 AMZN   Amazon    2024 2025-02-07 59248000000
 6 AMZN   Amazon    2025 2026-02-06 77670000000
 7 GOOGL  Alphabet  2020 2021-02-03 40269000000
 8 GOOGL  Alphabet  2021 2022-02-02 76033000000
 9 GOOGL  Alphabet  2022 2023-02-03 59972000000
10 GOOGL  Alphabet  2023 2024-01-31 73795000000
# ℹ 21 more rows

We’ll Join revenue + net income into one fundamentals table

fundamentals_clean <- revenue_clean %>%
  select(symbol, company, fy, revenue) %>%
  left_join(
    net_income_clean %>%
      select(symbol, fy, net_income),
    by = c("symbol", "fy")
  ) %>%
  mutate(
    revenue_billions = revenue / 1e9,
    net_income_billions = net_income / 1e9
  ) %>%
  arrange(symbol, fy)

fundamentals_clean

# A tibble: 31 × 7
   symbol company     fy revenue net_income revenue_billions net_income_billions
   <chr>  <chr>    <int>   <dbl>      <dbl>            <dbl>               <dbl>
 1 AMZN   Amazon    2020 3.86e11    2.13e10             386.               21.3 
 2 AMZN   Amazon    2021 4.70e11    3.34e10             470.               33.4 
 3 AMZN   Amazon    2022 5.14e11   -2.72e 9             514.               -2.72
 4 AMZN   Amazon    2023 5.75e11    3.04e10             575.               30.4 
 5 AMZN   Amazon    2024 6.38e11    5.92e10             638.               59.2 
 6 AMZN   Amazon    2025 7.17e11    7.77e10             717.               77.7 
 7 GOOGL  Alphabet  2020 1.83e11    4.03e10             183.               40.3 
 8 GOOGL  Alphabet  2021 2.58e11    7.60e10             258.               76.0 
 9 GOOGL  Alphabet  2022 2.83e11    6.00e10             283.               60.0 
10 GOOGL  Alphabet  2023 3.07e11    7.38e10             307.               73.8 
# ℹ 21 more rows

Now let’s calculate annual revenue and net income growth

fundamentals_growth <- fundamentals_clean %>%
  group_by(symbol, company) %>%
  arrange(fy) %>%
  mutate(
    revenue_growth_pct = (revenue / lag(revenue) - 1) * 100,
    net_income_growth_pct = (net_income / lag(net_income) - 1) * 100
  ) %>%
  ungroup() %>%
  rename(year = fy)

fundamentals_growth

# A tibble: 31 × 9
   symbol company   year revenue net_income revenue_billions net_income_billions
   <chr>  <chr>    <int>   <dbl>      <dbl>            <dbl>               <dbl>
 1 AMZN   Amazon    2020 3.86e11    2.13e10            386.                21.3 
 2 GOOGL  Alphabet  2020 1.83e11    4.03e10            183.                40.3 
 3 META   Meta      2020 8.60e10    2.91e10             86.0               29.1 
 4 MSFT   Microso…  2020 1.43e11    4.43e10            143.                44.3 
 5 NVDA   NVIDIA    2020 1.09e10    2.80e 9             10.9                2.80
 6 AMZN   Amazon    2021 4.70e11    3.34e10            470.                33.4 
 7 GOOGL  Alphabet  2021 2.58e11    7.60e10            258.                76.0 
 8 META   Meta      2021 1.18e11    3.94e10            118.                39.4 
 9 MSFT   Microso…  2021 1.68e11    6.13e10            168.                61.3 
10 NVDA   NVIDIA    2021 1.67e10    4.33e 9             16.7                4.33
# ℹ 21 more rows
# ℹ 2 more variables: revenue_growth_pct <dbl>, net_income_growth_pct <dbl>

Now we need to join fundamentals growth with stock growth

analysis_data <- annual_stock_growth %>%
  left_join(
    fundamentals_growth %>%
      select(
        symbol, company, year,
        revenue_billions, net_income_billions,
        revenue_growth_pct, net_income_growth_pct
      ),
    by = c("symbol", "year")
  ) %>%
  filter(!is.na(revenue_growth_pct)) %>%
  arrange(symbol, year)

analysis_data

# A tibble: 26 × 10
   symbol  year first_price last_price stock_growth_pct company revenue_billions
   <chr>  <dbl>       <dbl>      <dbl>            <dbl> <chr>              <dbl>
 1 AMZN    2021       159.       167.              4.64 Amazon              470.
 2 AMZN    2022       170.        84             -50.7  Amazon              514.
 3 AMZN    2023        85.8      152.             77.0  Amazon              575.
 4 AMZN    2024       150.       219.             46.3  Amazon              638.
 5 AMZN    2025       220.       231.              4.81 Amazon              717.
 6 GOOGL   2021        85.6      144.             67.8  Alphab…             258.
 7 GOOGL   2022       144.        87.5           -39.1  Alphab…             283.
 8 GOOGL   2023        88.4      139.             56.7  Alphab…             307.
 9 GOOGL   2024       137.       188.             37.5  Alphab…             350.
10 GOOGL   2025       189.       313.             65.9  Alphab…             403.
# ℹ 16 more rows
# ℹ 3 more variables: net_income_billions <dbl>, revenue_growth_pct <dbl>,
#   net_income_growth_pct <dbl>

Now that we have our dataset, let’s visualize the stock growth vs revenue growth.

ggplot(analysis_data, aes(x = revenue_growth_pct, y = stock_growth_pct, label = symbol)) +
  geom_point(size = 3) +
  geom_text(nudge_y = 5, check_overlap = TRUE) +
  geom_smooth(method = "lm", se = FALSE) +
  labs(
    title = "Stock Growth vs. Revenue Growth",
    subtitle = "Comparing annual stock returns against annual revenue growth for major AI-related companies",
    x = "Revenue Growth (%)",
    y = "Stock Growth (%)"
  ) +
  theme_minimal()

`geom_smooth()` using formula = 'y ~ x'

Warning: The following aesthetics were dropped during statistical transformation: label.
ℹ This can happen when ggplot fails to infer the correct grouping structure in
  the data.
ℹ Did you forget to specify a `group` aesthetic or to convert a numerical
  variable into a factor?

The scatterplot indicates a moderate positive relationship between revenue growth and stock performance, suggesting that stronger financial performance is often associated with stock appreciation. However, the variability in outcomes and presence of significant outliers imply that market valuations may also be influenced by investor sentiment and AI-related growth expectations beyond reported financial fundamentals.

Let’s further examine this with statistical analysis.

stock_model <- lm(
  stock_growth_pct ~ revenue_growth_pct + net_income_growth_pct,
  data = analysis_data
)

summary(stock_model)

Call:
lm(formula = stock_growth_pct ~ revenue_growth_pct + net_income_growth_pct, 
    data = analysis_data)

Residuals:
    Min      1Q  Median      3Q     Max 
-109.21  -42.25  -12.04   25.56  212.84 

Coefficients:
                       Estimate Std. Error t value Pr(>|t|)
(Intercept)           33.229755  21.211578   1.567    0.131
revenue_growth_pct     0.397526   0.531842   0.747    0.462
net_income_growth_pct  0.001077   0.061296   0.018    0.986

Residual standard error: 76.15 on 23 degrees of freedom
Multiple R-squared:  0.02989,   Adjusted R-squared:  -0.05447 
F-statistic: 0.3543 on 2 and 23 DF,  p-value: 0.7054

The regression analysis found no statistically significant relationship between annual stock growth and the financial performance metrics included in the model, specifically revenue growth and net income growth (p > 0.05). With an R-squared of approximately 3%, the model explains very little of the variation in stock returns, suggesting that stock market performance for major AI-focused companies may be driven by factors beyond traditional financial fundamentals, such as investor sentiment, future growth expectations, or broader market trends.

Next, I want to compare stock growth vs revenue growth by company over time.

summary_comparison <- analysis_data %>%
  group_by(symbol, company) %>%
  summarize(
    avg_stock_growth = mean(stock_growth_pct, na.rm = TRUE),
    avg_revenue_growth = mean(revenue_growth_pct, na.rm = TRUE),
    .groups = "drop"
  ) %>%
  pivot_longer(
    cols = c(avg_stock_growth, avg_revenue_growth),
    names_to = "metric",
    values_to = "growth_pct"
  ) %>%
  mutate(
    metric = recode(
      metric,
      avg_stock_growth = "Average Stock Growth",
      avg_revenue_growth = "Average Revenue Growth"
    )
  )

ggplot(summary_comparison, aes(x = reorder(symbol, growth_pct), y = growth_pct, fill = metric)) +
  geom_col(position = "dodge") +
  coord_flip() +
  labs(
    title = "Market Growth vs Financial Growth",
    subtitle = "Average annual stock returns compared with average annual revenue growth",
    x = "Company",
    y = "Average Growth (%)",
    fill = ""
  ) +
  theme_minimal()

Across all five AI-focused companies analyzed, average annual stock growth exceeded average annual revenue growth, with the largest gap observed in NVIDIA. This suggests that investor valuation growth has outpaced underlying business expansion, supporting the hypothesis that market expectations around AI may be contributing to elevated stock performance beyond traditional financial fundamentals.

I am going to run one more statistical test that directly supports the conclusion, because the regression showed “not significant.

Let’s do a simpler comparison:

Is average stock growth statistically higher than average revenue growth?

t.test(
  analysis_data$stock_growth_pct,
  analysis_data$revenue_growth_pct,
  paired = TRUE
)

    Paired t-test

data:  analysis_data$stock_growth_pct and analysis_data$revenue_growth_pct
t = 1.0811, df = 25, p-value = 0.29
alternative hypothesis: true mean difference is not equal to 0
95 percent confidence interval:
 -14.48624  46.49699
sample estimates:
mean difference 
       16.00538 

While the visual analysis suggests that stock market growth generally exceeds revenue growth among the AI-focused companies examined, the statistical tests did not find this difference to be significant at conventional thresholds (p > 0.05). This indicates a directional pattern in the data, but the relatively small sample size and variability across companies limit the strength of the conclusion. As a result, the findings suggest that investor valuations may be influenced by factors beyond traditional financial performance, though stronger evidence would require a larger dataset or broader company sample.

Let’s do the stock price over time:

ggplot(stock_prices, aes(x = date, y = adjusted, color = symbol)) +
  geom_line(size = 1) +
  labs(
    title = "Stock Price Trends for Major AI-Focused Companies",
    subtitle = "Adjusted closing prices from 2020 to present",
    x = "Date",
    y = "Adjusted Stock Price ($)",
    color = "Company"
  ) +
  theme_minimal()

Warning: Using `size` aesthetic for lines was deprecated in ggplot2 3.4.0.
ℹ Please use `linewidth` instead.

The stock price trend visualization illustrates the market performance trajectories of major AI-focused technology companies from 2020 onward. While all companies experienced growth over the study period, the magnitude and volatility of performance varied substantially, with NVIDIA showing particularly sharp acceleration in recent years. This visualization provides descriptive context for the broader analysis by demonstrating how investor valuations evolved over time prior to comparison with underlying financial fundamentals.

Executive Summary

summary_table <- analysis_data %>%
  group_by(symbol, company) %>%
  summarize(
    avg_stock_growth = mean(stock_growth_pct, na.rm = TRUE),
    avg_revenue_growth = mean(revenue_growth_pct, na.rm = TRUE),
    .groups = "drop"
  ) %>%
  mutate(
    stock_minus_revenue = avg_stock_growth - avg_revenue_growth
  )

summary_table

# A tibble: 5 × 5
  symbol company   avg_stock_growth avg_revenue_growth stock_minus_revenue
  <chr>  <chr>                <dbl>              <dbl>               <dbl>
1 AMZN   Amazon                16.4               13.3                3.16
2 GOOGL  Alphabet              37.8               17.7               20.0 
3 META   Meta                  44.9               19.2               25.7 
4 MSFT   Microsoft             23.5               14.6                8.87
5 NVDA   NVIDIA                91.1               70.0               21.2 

Across all five AI-focused companies analyzed, average annual stock growth exceeded average annual revenue growth. The largest gaps were observed for Meta, NVIDIA, and Alphabet, where market appreciation substantially outpaced business growth. While earlier statistical tests did not find this relationship to be statistically significant, the directional consistency across all companies suggests that investor expectations surrounding AI may be contributing to elevated stock valuations beyond traditional financial fundamentals.