Tag Archives: aws

Y Google Cloud?

Posted on by

I spent the last year working on fine-tuning and evaluating Gemini (Google’s version of ChatGPT) inside of Google Cloud. While I had a great experience learning what it takes to bring a model like this into production, after a few months inside Cloud, I struggled to understand both why Google was investing so heavily in this business and why Wall Street was so interested. I know that playing armchair CEO is easy and running a Fortune 5 company is hard, but the following is my argument against such a concentrated (~30% of Google’s headcount + capex) bet on Cloud.

Note: these prioritization discussions can’t take place in a vacuum. My take is not that Google Cloud is bad in isolation but rather that the opportunity cost to invest in cloud is high and Google would be better suited directing some or all of these resources into consumer products more aligned with Google’s comparative advantages.

TL;DR

Google’s core businesses are quite strong, still growing reasonably fast, and capable of accelerating with additional resources that could be redirected away from Cloud for the following reasons:

Great reasons to invest in Cloud (and why these don’t apply to Google)

Excess compute capacity

Apparently, the commonly repeated AWS founding story that they were leasing out excess capacity isn’t true, but if a company does have excess compute or storage capacity, leasing it out to 3P is a great idea. However, Google, like many other tech companies, does not have that problem. Every team at Google needs to fight for chips to serve fancier models, run more experiments, add features, and so on. Every FLOP that is leased to Cloud customers could almost certainly be leveraged more efficiently by an internal team.

Worse margins in core business

According to their most recent earnings report, last year Amazon retail earned $6B on $146B in sales netting a lovely 4% margin. In contrast, AWS earned $7B on $24B coming in at 29%, a >7x improvement.

Google on the other hand has a great core business. Last quarter, they earned $28B on $76B in revenue (40% margin) on advertising, services, and devices and only $1B on $9B (11% margin) on Cloud (note that the Cloud expenses are likely optimistic given that significant core Google resources support Cloud products). This means that Google needs to sell almost 4x more Cloud services than advertising to have the same impact on market cap (assuming a constant P/E).

Core business ceases to grow (or begins to shrink)

Both Amazon and Microsoft, the #1 and #2 cloud computing providers, were facing existential threats to their core businesses when they began investing in Cloud. After nearly two decades of operation, Amazon had to turn a profit someday, but the margins in online retail proved to make that nearly impossible. Microsoft had to replace the revenue of their OS business with something as the era of the personal computer and Windows server came to a close.

In contrast, Google’s core search business still achieves impressive margins and grew 13% last year. Its “Google subscriptions, platforms, and devices” division grew 23% (while still generating $11B in revenue!), which is comparable to Cloud, more aligned with Google’s comparative advantages, and likely higher margin.

Hard to find scale

Companies like Google are large enough that new business ventures must have the potential of generating O($10B/year). There are relatively few sectors that support this scale and Cloud is one of them ($563B in 2023).

However, Cloud is not the only one. Apple sold $70B in phones last quarter, Meta $40B in ads, Uber $31B in rides, and so on. And in each of these areas, Google has invested years in building competitive products. If I were Sundar and wanted to find an extra $9B in revenue that would grow at 26% y/y, I think that I could do that more easily by taking a bigger chunk out of the devices market, building new surfaces for online advertising as ATT and other macro trends squeeze the marginal players, and pushing Waymo to expand to more markets over creating a Cloud business where it’s not obvious how Google’s product is differentiated or aligned with its comparative advantage.

Diversification

I don’t actually buy this argument but wanted to include it because the importance of diversification is commonly cited as justification for Google Cloud. I disagree with this assertion because:

  • the top companies in the world are quite concentrated in what they offer
  • Google is already diversified into quite a few businesses (search ads, video ads, as network, devices, services, autonomous driving, healthcare, etc.) that could all grow with adequate resources

Starting with my first point, the top 10 public companies in the world by market cap today (02/19/24) are:

  • Microsoft
  • Apple
  • Saudi Aramco
  • Nvidia
  • Alphabet
  • Amazon
  • Meta
  • Berkshire Hathaway
  • Eli Lilly
  • TSMC
  • Broadcom
  • Tesla
  • Visa
  • Novo Nordisk
  • JPMorgan Chase

I would only consider 3 of these to be more diversified than Google today without Cloud: Berkshire Hathaway, Amazon, and Microsoft. Each of the others found a niche and grew enormously without making large shifts. I don’t think that anyone is telling Elon Musk to divert resources away from making cars or Jensen that he is too fixated on GPUs.

The opportunity cost of Google’s investment in Cloud

Headcount

Google Cloud employs 54k employees, which represents 30% of Alphabet’s headcount. These employees generate 1% of Google’s profits while achieving a growth rate <2x the 25 year-old search business and on par with “subscriptions, platforms, and devices”. And this is even worse than it sounds. Because the Cloud margins are almost 4x worse than these other businesses, from an earnings perspective, Cloud is actually growing more slowly.

Novel employee profiles

Outside of Cloud, Google employees are quite fungible. A SWE or designer on search can switch to YouTube and feel right at home within a few weeks. Consumer internet products are all roughly the same shape in terms of the skillset they demand. Cloud is very different. To sell cloud products, you need OPMs (Outbound Product Managers), CEs (Customer Engineers), and an alphabet soup of other roles that don’t really exist elsewhere inside of Google.

The consequence is twofold:

  • Google incurs the overhead of maintaining more career ladders and HR systems for recruiting and retention
  • The internal mobility of these employees is limited to Cloud, likely reducing retention and satisfaction

Missing innovation in core products

Life is about tradeoffs. These 54k people working on Cloud are not working on Google’s core products like Search, Maps, Android, Pixel, and so on. Several years ago, I would have argued that this would not have been a problem and that Google had a comfortable lead in many of these areas. This is no longer true.

AI upstarts like OpenAI, Arc Browser, and Perplexity are almost certainly siphoning away search volume (and could potential disrupt web search entirely), Apple Maps has emerged as a formidable competitor to Google Maps, the green text bubbles have pushed the majority of Americans to iPhone, TikTok is stealing eyeballs from YouTube, and so on.

I believe that with more headcount, chips, exec attention, and other limited resources, Google could reverse the advances competitors have made and resolidify the lead that all of their core products once held.

Cloud is undifferentiated

At the end of the day, every cloud provider is basically selling the same thing: reliable compute and storage. Yes, there are bells and whistles and AI services, but the bread and butter is offering enterprises the opportunity to move their workloads out of their data centers and into yours most easily and cheaply. While Google has historically competed by building delightful products that billions of consumers have learned to love, Cloud largely differentiates through investments in customer support, sales, and marketing.