Welcome back to Funnel Vision, the weekly newsletter that puts performance creative & growth marketing in focus. Brought to you by the crew at Ready Set.

Five gigawatts is enough electricity to power one million average homes continually for an entire year. It’s also the amount of power Meta’s largest AI data center will consume once it’s fully operational.

Here’s what’s inside today’s edition:

• A deep dive into the story behind Meta’s gangbusters Q2 earnings

• What it means for advertisers, brands, and society at large

Reading time: 6 minutes

🌴 Quick note: Funnel Vision is going on summer vacation! We'll be back to regularly scheduled programming August 22.

AD INDUSTRY NEWS

How Meta’s Roaring Ad Business Is Funding the AI End Game

ICYMI: Meta had a massive Q2. Earnings blew expectations out of the water, which combined with a raised Q3 forecast fueled a 10%+ stock pop.

Yet hiding under the impressive performance numbers loomed an even bigger and more important story: Meta’s plans for AI domination. 

Because we’re major nerds for stuff like this, we tuned into the earnings call and read all of the press releases, presentations, and financial charts. Yes, we understand such activities aren’t high on the to-do list for most normal people busy marketers.

So we distilled everything down to the most essential takeaways.

First, a quick snapshot of the vitals from the earnings report: 

• Overall revenue for Q2 was $47.5B (vs. $44.8B expected) a 22% YoY increase.

• Operating income/margin: $20.4B/43% 

• Total ad revenue was $46.6B, also up 22%. This is 98% of overall revenue.

• Ad impressions up 11% and average price per ad up 9%.

• Video engagement up 20% YoY on both Instagram and Facebook.

• Guidance for Q3 raised to $47.5–$50.5B, above estimates of $46.1B.

Now let’s take a look at the bigger picture.

Meta Is All-In On AI

Mark Zuckerberg wasted no time diving straight into the topic on everyone’s mind: artificial intelligence.

• The headline: Superintelligence, which Zuckerberg defines as AI that surpasses human intelligence in every way, “is now in sight”.

• The vision: Bring personal superintelligence to everyone.

Meta’s been making waves in recent months with their… unconventional approach to acquiring talent. Whether through investment/acquihires or luring leading AI engineers and researchers with 9-figure pay packages, Zuckerberg has seemingly made it his personal mission to build the most “elite, talent-dense team” in AI.

(Side note: AI researchers earning NBA superstar-level salary packets was not on my 2025 bingo card.)

This hiring spree was so costly that Meta CFO, Susan Li, cited employee compensation as the second-largest expense for the remainder of 2025 and 2026, behind only…

AI Infrastructure Will Dominate CapEx

If assembling the ultimate team of AI avengers is one side of the superintelligence coin, then compute is the other. Having one without the other is like having Batman without the Batcave and all the shiny toys inside it.

The compute for the new Meta Superintelligence Lab will be supplied by massive multi-gigawatt clusters—huge data warehouses designed to power AI systems. Prometheus and Hyperion have been announced, with the latter expected to be the size of Manhattan and capable of outputting 5GW of compute power at full capacity.

(For scale, 5GW could power 4 million homes continuously for 1 year.)

Overall, Meta plans to spend $72B on AI infrastructure in 2025. Which raises a key question…

How Are They Going To Pay For It?

Okay, so Meta is building superintelligence. Cool. But how does that affect us performance marketers?

Because they plan to pay for it with their booming ad business.

(Oversimplification? Yes. On the right track? Also yes.)

Before we collectively clutch our pearls, this is a good thing. As noted above, ad revenue accounts for 98% of Meta’s overall revenue. If they want the cash to pay for these eye-watering AI investments, keeping advertisers happy—and paying the bills—is paramount.

Improvements to advertising tools fall into three major buckets: systems, products, and platforms.

Improvements To Ad Systems

Meta is continuing to optimize the ad retrieval and ranking engine to better serve more relevant and personalized ads. The primary pillars of this engine are characters we’ve met before:

Andromeda: The ads retrieval system that whittles down tens of millions of potential candidates to the few thousand most relevant ads. Improvements have driven 4% higher conversions on Facebook.

GEM: GEM powers the ranking stage, the step after retrieval, that determines the best ads to show from Andromeda’s shortlist. Scaled training capacity, more engagement data, and doubled event sequences have driven 3-5% increased ad conversions on Instagram and Facebook.

Lattice: Expanded coverage of the Lattice model architecture—the system that consolidates learnings across objectives, making for better optimization—has resulted in a nearly 4% increase in ad conversions.

In short: Major upgrades to Meta’s ad systems are beginning to materialize.

Improvements to Ad Products

Upgrades to core ad products, like the Advantage+ suite, and up-and-coming features, like Gen AI, are making Meta’s ad tools more powerful and easier to use.

Nearly two million advertisers are now using video generation features, which are part of a broader suite of AI tools Zuckerberg frames as being “especially valuable for smaller advertisers with limited budgets”. With larger advertisers, Meta expects agencies (like Ready Set) will continue to be valuable partners to “help larger brands apply these tools strategically” to drive performance.

Improvements to Ad Platforms

A couple notable upgrades on platform and measurement tools:

• Incremental attribution is now fully rolled out.

• Omnichannel ads also went live

• New, cheap ad inventory is also available on Threads and WhatsApp

Meta’s Big Bold Bet… AI Glasses

After several notable false starts and fumbles in the AR/VR hardware arena, it appears Meta may actually have a business case for the years of work and billions of dollars they’ve poured into wearables.

Zuckerberg views AI glasses as the “main way we integrate superintelligence day-to-day”.

The market seems to agree. Sales of Ray-Ban Meta glasses are accelerating, with demand outstripping supply. A new partnership with Oakley (complete with an action sports star-packed hype video) introduces several new styles. 

It’ll be interesting to see how this pans out. If Zuckerberg’s vision comes into focus, and glasses really are the primary human/AI interface, Meta is already several years ahead on building this tech.

One major caveat: Reality Labs, the division responsible for developing the glasses, keeps losing money ($4.5B operating loss in Q2 😬).

What It All Means

Meta is investing heavily in the race to AI superintelligence because they believe they can win.

Thanks to their exceptional ad business, which rakes in almost $50B every three months, they’re also one of the only companies in the world capable of assembling both the talent and the compute required to do it.

The boon to advertisers? The ad business essentially pays for Meta’s AI investments. Meaning the focus is very much on keeping customers (i.e. advertisers) happy.

This quote from Zuckerberg on the earnings call (edited for clarity) nicely summarizes the vision for the future:

FROM OUR SPONSOR

Thanks for reading!

That’s all for this week. Seeya next time 🫶 

Psst... wanna earn $1k?

Know someone looking for efficiency, flexibility, and scale with performance creative and media buying? Refer brands in your network and earn $1,000.

Check out the Ready Set referral program here.