Earnings Rewind: AI Gold Rush Meets Reality, LatAm Fintech Soars, and Garena’s Big Comeback

Earnings Rewind: AI Gold Rush Meets Reality

Welcome back to Earnings Rewind, your weekly breakdown of the market’s most important earnings calls that I didn’t cover in another article. This week, we saw a stark divide between the digital and the physical. On one side, digital-first international players Sea Limited and Nu Holdings posted stellar results, proving that their platforms are scaling with impressive profitability.

On the other side, the physical tech companies showed that building the future is difficult, expensive, and subject to the messy realities of the physical world. AI infrastructure provider CoreWeave announced a an impressive $55 billion backlog… and then promptly cut its guidance due to construction delays. And Rocket Lab, a beacon of execution in the space industry, also beat its numbers but had to push the debut of its flagship rocket, Neutron, into 2026.

CoreWeave: The $55 Billion Backlog Weighed Down by a Bottleneck

AI cloud provider CoreWeave (NASDAQ: CRWV) reported its Q3 2025 earnings on Monday, and it was a little bit of good news, bad news. The demand for its specialized AI cloud is so unfathomably high that the company can’t build data centers fast enough—literally.

Stock Performance: The market was mostly focused on the bad news. After reporting on Monday, the stock, which had been a high-flyer since its IPO, went into a tailspin falling 30% in the past five days. The stock is still up 93% YTD but that mostly came in the first few months post IPO so any buyers after May are now down. This wasn’t just a Coreweave issue as many AI plays especially the more speculative ones were down a lot in the past week or two(see competitor NBIS as well).

The Earnings: The numbers were a paradox.

• Revenue: $1.4 billion, up a staggering 134% year-over-year and beating expectations.
• Revenue Backlog: This is the headline. The backlog almost doubled in a single quarter, ending at $55.6 billion. The company added over $25 billion in backlog in Q3 alone. That really shows the huge demand for what Coreweave offers.
• The Guidance Cut: Despite the insatiable demand, CoreWeave cut its full-year 2025 revenue guidance to ~$5.1 billion (from a previous ~$5.3B) and slashed its 2025 CapEx forecast by 40%. The main issue is that they have a substantial amount of construction in progress so the CapEx(mostly GPU deployment) is likely just being pushed out to early 2026 versus late 2025.

Pros: Insatiable Customer Demand

• Diversification: The company has brilliantly diversified its customer base. At the start of the year, one customer (rumored to be Microsoft) represented 85% of its backlog. Today, no single customer is more than 35%.
• Blue-Chip Wins: The backlog is filled with the biggest names in tech, including large-scale compute contracts with Meta and OpenAI. It also signed its sixth contract with an unnamed leading hyperscaler and onboarded new enterprise customers like CrowdStrike and NASA.
• Tech Leadership: CoreWeave continues to prove it’s a big player in an increasingly crowded space. It was first-to-market with the GB300s and was the only cloud provider to submit MLPerf Inference results for them. It also received the Platinum ClusterMAX ranking from SemiAnalysis for the second time, ahead of all hyperscalers.

Cons: The Physical World Bites Back

• Supply Chain Bottleneck: The guidance cut was entirely due to temporary delays related to a third-party data center developer who is behind schedule. This isn’t a problem with GPU supply (like last year) but with the physical powered-shell buildings.
• Guidance and CapEx Cut: A 40% cut to CapEx and a revenue guidance reduction, no matter the excuse, is a terrible look. It spooked investors and raised questions about the lumpy nature of building out AI infrastructure at this scale and potentially worried them about actually being able to deliver on those large contracts they’ve signed.
• Profitability: Building out this infrastructure doesn’t come cheap. Coreweave already has $17B in net debt which is a substantial part of its $55B enterprise value. While their growth is impressive and expected to continue for years and years, so is their free cash flow burn. It’s quite like that they’ll be throwing off $13-$15 in free cash flow per year for at least the next two years and likely won’t be self funding until 2029 at the earliest. That’s a lot of additional debt or dilution that investors will have to swallow and a lot of dependence on the demand/supply dynamic favoring demand for a while.

From the Q&A: Analysts hammered management on the delay. CEO Mike Intrator was quick to frame it as a systemic problem for the entire industry, not just CoreWeave. He stressed that the delay was from a single data center provider (out of 32 in their portfolio) and, crucially, the affected customer agreed to adjust the delivery schedule and extend the expiration date, preserving the full value of the contract.

Analysts also asked if this meant the infrastructure was locked to one customer. Management confirmed the infrastructure is fungible and could be used for any customer. The CFO also clarified that the CapEx cut wasn’t a cancellation, but a shift, with the vast majority of the spending now landing in Q1 2026.

Overall, Coreweave’s earnings delivered more of the same but it’s clear that investors are being a bit more wary about the long term commitment of companies such as this one and Nebius and the investments that will need to be made to get them to free cash flow positivity. If the AI boom starts to slow a bit, these companies could be in trouble so speculative names like those two get punished as investors take a more risk-off approach.

Sea Limited: Garena’s Roaring Comeback Powers a Full-Spectrum Beat

Sea Limited (NYSE: SE) reported on Tuesday, and it was the reverse of its 2023-2024 narrative. After a painful period of focusing on profitability over growth, Sea proved it can now do both. All three of its business units fired on all cylinders, with its long-dormant gaming division, Garena, posting a spectacular comeback.

Stock Performance: The market’s reaction was perplexing. Despite a massive revenue beat and raised guidance, the stock fell around 7.6% in the past 5 days. The likely culprit? A miss on net income. Sea is spending money again, and Wall Street, still scarred from the last down-cycle, got nervous.

The Earnings: The quarter was a show of force.

• Revenue: $6 billion, up 38% year-over-year.
• Adjusted EBITDA and Operating Income: $874 million, up 68% year-over-year and Operating Income was up 135.1%.
• Shopee (E-Commerce): GMV grew 28% YoY, its fifth straight quarter of sequential growth. The company raised its full-year GMV growth guidance from 20% to over 25%. Revenue was up 34.9%
• Garena (Gaming): The star of the show. Bookings surged 51% YoY to $841 million, its best quarter since 2021. This is big as this is a higher margin business that can fund investments in other areas of the business.
• Monee (Fintech): The loan book grew 70% YoY to $7.9 billion, while the 90-day non-performing loan (NPL) ratio remained a healthy 1.1%.

Pros: The Three-Headed Monster is Back

• Garena’s Resurrection: Management credited blockbuster IP collaborations for the gaming division’s turnaround. Events for Squid Game (300 million+ plays) and Naruto Shippuden Chapter 2 were massive successes, driving both engagement and revenue. The company raised Garena’s full-year bookings growth guidance to over 30%.
• Shopee’s Logistics Moat: CEO Forrest Li detailed how SPX Express is a key differentiator. In Indonesia, it offers same-day delivery in cities and economical (20% cheaper) delivery in rural areas. In Taiwan, its 2,500+ self-pickup lockers handle 70% of deliveries at a 30% lower cost. The next step: building a fulfillment (warehouse) capability to deepen their logistics competitive moat.
• Fintech Flywheel: Monee is scaling rapidly, adding 5 million first-time borrowers in Q3. The all-can-apply approach for Shopee users, paired with instant credit approvals, is successfully converting e-commerce users into high-margin financial services customers.

Cons: Investing in Growth (and Spooking the Market)

• The Profit “Miss”: Net income missed analyst estimates because Sea is spending again. E-commerce adjusted EBITDA margins dipped quarter-over-quarter as the company invested in their fulfillment capacities but growth was still north of 100% for the quarter.
• Intentional Spending: This spending is targeted. The company is investing heavily in its logistics and fulfillment network and its new Shopee VIP subscription program, which already has 3.5 million members. Management noted VIP members spend ~40% more after joining.

From the Q&A: The first question from analysts was, predictably, about the margin dip in e-commerce. Management’s answer was clear: this is an intentional investment, not a return to the cash-burning days. This investment cycle is quite different from two years ago, as it’s less about marketing and more about continuous investment… to strengthen our competitive moat, which also drives growth.

When asked about competition from live-streaming peers (like TikTok), management was unconcerned, stating the landscape is relatively stable and that a content-driven platform is fundamentally different from Shopee’s shelf-based marketplace.

Overall, these were really solid earnings. Investors will keep a close eye whether or not the gaming division can sustain growth or if it this is a one off year due to very successful events making for a tougher comparable next year. If it can then that higher margin business can really offset the investments being made in e-commerce.

Nu Holdings: The AI-First Fintech Flywheel Hits 127 Million Customers

Nu Holdings (NYSE: NU) continued its reign as one of the world’s most successful fintechs, delivering another flawless quarter on Thursday. The company smashed estimates, raised its ROE, and revealed a surprisingly advanced AI strategy that sets it apart from its peers.

Stock Performance: In a rough week for tech, Nu didn’t do too terribley. The stock hit a new all-time high of $16.39 on Tuesday following the report but dive give back some of those gains and ended the week down 2% as the stock market was soft overall.

The Earnings: Just another record-breaking quarter.

• Customers: Grew to 127 million (adding 4 million in Q3).
• Revenue: A record $4.2 billion.
• Profitability: A record net income of $783 million and a record ROE of 31%.
• Efficiency: Cost-to-income ratio fell to a new low of 27.7%.

Pros: The Model is Working Perfectly

• The Mexico Story: The growth story is all about Mexico. Nu surpassed 13 million customers there, reaching 14% of the adult population. Incredibly, its ARPAC (Average Revenue Per Active Customer) in Mexico is already $12.5, nearly matching Brazil’s $13.4, despite not having a full suite of products yet. This market is growing much faster than Brazil and has a lot more potential ahead of it it’s its maturity is anywhere close to Brazil.
• Brazil (The Cash Cow): The mature Brazil market is a fortress, serving over 60% of the adult population and funding the international expansion.
• The nuFormer AI Engine: CEO David Vélez unveiled a new AI-first strategy. Nu has developed its own proprietary foundation model called nuFormer, with 330 million parameters trained on 600 billion tokens of its own data. This is already being used to improve credit models, allowing Nu to meaningfully increase limits for eligible customers while maintaining the same overall risk appetite. Whether or not this will provide anything remains to be seen but the market generally loves any and all AI buzzwords so why not.
• Stellar Credit Quality: The biggest positive surprise was a 7% decline in Credit Loss Allowance expenses. This wasn’t a fluke; it was driven by better underwriting (maybe partially thanks to AI) and better collection strategies. The all-important risk-adjusted net interest margin (NIM) expanded from 9.2% to 9.9%.

Cons: Not Too Many

• Headline NIM Contraction: The headline NIM contracted by 40 basis points. However, this was an intentional and positive strategic move, reflecting a shift toward lower-risk, lower-yield products like secured lending.
• Regulatory Risk: The company faces a potential headwind in Mexico from a proposed cap on credit card interchange fees.

From the Q&A: Analysts immediately grilled management on the surprisingly low provisions (cost of risk). The CFO explained this was due to better-than-expected portfolio performance, a new reactivation program for defaulted customers that led to higher recoveries, and the new AI-driven credit models.

When pressed on Mexico’s profitability, David Vélez gave a blunt and powerful answer: “If we wanted to be profitable in Mexico, we would be profitable already… That would actually be a really bad decision.” He was clear that the company is optimizing for long-term growth and market share, not short-term profit, in its new high-growth markets.

Overall, these earnings were very impressive. This might be on I have to start a small position in as the growth story in Mexico is very impressive.

Rocket Lab: Record Growth and a $1B War Chest, But Neutron Will Have to Wait

Rocket Lab (NASDAQ: RKLB) reported on Monday, delivering a quarter that perfectly captured the company’s dual nature: its existing Electron rocket business is firing on all cylinders, while its ambitious next-gen rocket, Neutron, is proving to be incredibly difficult.

Stock Performance: The market, which hates delays above all else, punished the stock. Shares fell around 15% for the week.

The Earnings: The numbers themselves were strong.

• Record Revenue: $155 million, up 48% year-over-year and at the high end of guidance.
• Record Electron Bookings: The workhorse Electron rocket had its best quarter ever, with 17 dedicated launches signed. The launch backlog now sits at 49 missions.
• Strong Guidance: Q4 revenue is guided to $170M-$180M, another sequential jump.

Pros: The End-to-End Moat is Widening

• $1 Billion M&A War Chest: Rocket Lab raised $468.8 million during the quarter, bringing its total liquidity to over $1 billion. Management explicitly called this dry powder for future M&A, with the CFO noting the pipeline is active after recently acquiring Geost for $325m.
• Space Systems is the Star: The satellite-building division was the main growth driver, with revenue up 16.7% sequentially. Its two spacecraft for NASA’s ESCAPADE mission to Mars are integrated and ready for their upcoming launch.
• HASTE is a Hit: The hypersonic test vehicle, HASTE (an Electron variant), had back-to-back successful missions and is seeing strong demand from defense customers.
• Acquisitions Paying Off: The recent GEOs acquisition is successfully opening doors with the intelligence community, allowing Rocket Lab to compete as a prime contractor.

Cons: Rockets Are Hard

• The Neutron Delay: The one con was a big one. The first launch of the next-generation, reusable Neutron rocket is being pushed from late 2025 to early next year” (2026).
• Government Shutdown Woes: The ongoing government shutdown has delayed the award of the massive SDA Tranche-3 constellation contract, a major opportunity for the Space Systems division.

From the Q&A: CEO Peter Beck was defiant about the Neutron delay, framing it as a strategic choice, not a failure. He explained the company is sticking to its meticulous Rocket Lab process to retire the risks on the ground first. In a clear shot at competitors who have had rockets explode on the pad, Beck said, “You won’t see us… just clearing the pad and claim[ing] success… We’re seeing what happens when others rush to the pad with an unproven product, and we just refused to do that.”

That sounds great but hopefully they don’t actually fail whenever that actually launches.

When asked if the delay hurts their chances for big government launch contracts (NSSL), management was unconcerned, noting those awards were always contingent on a successful flight first. The delay doesn’t change that, it just pushes out the timeline.

Rocketlab is a pricy stock but they are in an industry that has maybe 3 major players. A $24B market cap against $600m in 25 estimated revenue is really dependent on hyper growth for a while so it’s no surprise to see the stock punished for any minor mistakes. The stock is still up 140% in the past 12 months so investors should be happy.

Conclusion: The Digital vs. The Physical

This week was a perfect illustration of two different worlds. In the digital realm, Nu Holdings and Sea Limited showed the power of mature, scaled platforms. They are now flywheels, using their massive customer bases and data advantages (like Nu’s nuFormer AI) to drive new growth and accelerating profits.

In the physical world, CoreWeave and Rocket Lab reminded us that building things is hard. Even with a $55 billion backlog or a $1 billion war chest, you are still at the mercy of third-party construction crews and the unforgiving laws of physics. For these hard tech champions, the demand is clearly huge; the only question is whether they can execute fast enough to meet it and whether the market is pricing them way too aggressively based on that demand. That’s why stocks like that can easily drop 15-30% in one week based on some bad news.

Disclosure: This is not investment advice. Please talk to a qualified financial advisor before making any investment decisions.