Inside Sea Group: Ecommerce, Gaming, Payments and AI

Why does Sea Group matter in 2026?

Sea Limited closed 2025 with $22.9 billion in revenue, $1.6 billion in net income, and three businesses operating at scale that almost no other company in the world can match — a regional e-commerce leader, a top-five global mobile game, and a digital bank with a $9.2 billion loan book. Net income was up 260% year-on-year. All three segments — Shopee, Garena, and the newly rebranded Monee — posted positive adjusted EBITDA. This is the company that nearly went bust in 2022 controlling roughly 52% of Southeast Asian platform e-commerce GMV and signing an MOU with Google to co-architect agentic AI infrastructure for the region.

If you want to understand where Southeast Asia’s digital economy is going, you have to start here. Sea isn’t just a regional success story anymore — it’s a stress test for whether a single company can hold three flywheels turning at once. The 2025 results suggest yes. The Q4 stock reaction — down 16% on results day — suggests the market isn’t fully convinced.

The deeper read, and the one most investors are still missing: Sea owns multiple context layers — gaming graph, payments rails, content history, purchase data — across the region’s densest super-app footprint. In an AI-agent world, context is the new moat. Whoever owns the persistent state of the user becomes the layer agents have to negotiate with. Single-layer aggregators get squeezed. Context-owners get re-rated up.

This piece walks through the three businesses, the strategic shifts that defined 2025, and what’s actually at stake in 2026.

Shopee vs Lazada vs Tokopedia: SEA’s $159B War

How big is Sea Group, really?

A clean snapshot of FY2025:

• Total revenue: $22.94 billion (+36% YoY)
• Net income: $1.58 billion (+260% YoY)
• Shopee GMV: record annual figure, $32.2 billion in Q3 2025 alone (+28% YoY)
• Garena bookings: $2.9 billion (+24% YoY)
• Monee loan book: $9.2 billion principal outstanding (+80% YoY)
• Shopee adjusted EBITDA: $881 million (vs $156 million in 2024)
• Monee adjusted EBITDA: $1 billion (+43% YoY)
• Market cap (early May 2026): approximately $52.8 billion at $86 per share

Sea trades on the NYSE as SE. The stock peaked above $199 in late 2025 and sold off sharply after Q4 results in March 2026, when Q4 EPS came in at $0.63 against a $0.80 consensus expectation. Analyst consensus 12-month price targets range from $137 to $195, with a 29-analyst average around $137-$174 depending on the day.

The numbers matter. The structural question matters more. How are the three businesses actually feeding each other — and which AI-era reality are they being built for?

Garena: Free Fire isn’t a game, it’s a place

Garena is the part of Sea that nearly killed the company. In 2022, India banned Free Fire over data security concerns, wiping out a material chunk of Garena’s user base and triggering a stock collapse from over $300 to under $50. Sea later settled a $40 million securities class action in August 2025 over allegations it had concealed deteriorating intra-quarter user metrics ahead of that ban. A separate $46 million settlement is pending for a different class period covering 2022-2023. Real money, real lessons.

What’s happened since is a textbook live-ops resurgence. Garena bookings grew 37% in 2024 and another 24% in 2025, hitting $2.9 billion. In Q1 2025 alone, bookings were up 51% year-on-year. Free Fire averaged over 100 million daily active players through 2025 — an eight-year-old game posting its best numbers since 2021.

But the more important framing is this: Free Fire isn’t a game. It’s a place. We don’t call TV viewers “watchers” or label social media users by the medium. Yet we still talk about “gamers” as if Free Fire’s 100 million daily players were consumers of a product rather than a population inhabiting an environment. In Indonesia, 83% of internet users play games. Philippines 82%. Vietnam 81%. Singapore 79%. Malaysia 75%. In five of seven core SEA markets, four out of every five internet users play games. Esports awareness in Vietnam runs at 94%, with 59% watching regularly. Mobile Legends: Bang Bang was a medal event at the 2025 SEA Games. These environments don’t behave like products. They behave like places — communities evolve whether you’re logged in or not, economies adjust, live ops follow Lunar New Year, Songkran, Eid, Ramadan.

That reframing matters because it changes what kind of asset Garena owns. The mechanism of Garena’s 2025 resurgence was collaboration-led live ops — Naruto Shippuden in January, Squid Game in July, Naruto Chapter 2 in late July, Gintama in April 2026 — each engineered to drive a spike in spending per user. The Squid Game crossover alone helped push Free Fire to sixth place in US mobile game sales in July 2025, up seven spots, and made it the highest-earning mobile shooter in the US (Circana data, reported by Bloomberg). Garena landed in the home market of Call of Duty Mobile and PUBG Mobile and beat both on revenue.

The strategic shift is sharper than “live ops” captures. Garena has effectively given up on being a third-party publisher. The Riot Games partnership — which had Garena publishing League of Legends and Teamfight Tactics across Southeast Asia for 12 years — ended in January 2023. Tencent titles still run, but the centre of gravity has moved entirely to Free Fire as a perpetual environment and to internal development. India remains the unfinished story: Garena announced a Free Fire India relaunch in mid-2025 with localised servers via Yotta Data Services, parental controls, and a ₹1 crore esports tournament, but as of late 2025 the game was still in pre-registration. A full relaunch in 2026 reopens a market with 8.45 billion mobile game downloads in FY2024-25 (Sensor Tower) — a meaningful catalyst nobody has properly priced.

Garena’s adjusted EBITDA grew faster than bookings in Q4 2025 — up 25.5% with margin expansion — which is what live ops at scale is supposed to deliver. Forrest Li, Sea’s CEO, has guided to double-digit bookings growth for 2026.

There’s a further optionality the market currently values at zero. Persistent multiplayer environments with physics, consequence and feedback loops are exactly what the next AI paradigm — world models, embodied agents — needs as training substrate. DeepMind, NVIDIA and World Labs are explicitly building toward this. Free Fire sessions, with billions of hours of human decision-making across diverse cultural contexts, are a structurally valuable asset for that paradigm. Garena is priced as a one-game company. The asset is more than that.

How Gaming Monetisation Works in Asia

Shopee: dominance, with credible challengers

Shopee held approximately 52% of Southeast Asian platform e-commerce GMV in 2024, and held that share through 2025 even as the regional market grew 22.8% to $157.6 billion (Momentum Works, 2026 SEA E-commerce Report). The platform GMV figure for 2024 was $66.8 billion. Of the $14 billion in regional GMV growth from 2024, Shopee captured roughly $12 billion. It’s the closest thing to a dominant regional consumer platform that Southeast Asia has ever produced.

But the gap is closing. TikTok Shop, including Tokopedia, reached 65.7% of Shopee’s GMV by the end of 2025, with the monthly gap narrowing further toward year-end. In Vietnam specifically, Shopee’s market share fell to 56% in 2025 from 61% in 2023, while TikTok Shop expanded to over 41% (Metric, 2025). The two platforms together account for 97% of Vietnam’s online platform GMV. Lazada, after years of bleeding share, posted its first full year of profitability in 2024 and has pivoted to a brand-led, higher-quality positioning rather than directly competing on price.

Shopee’s defence is built on three things: SPX Express logistics, advertising, and the Shopee VIP loyalty programme. SPX now fulfils more than half of all Shopee orders. Shopee VIP membership grew to over 3.5 million members by Q3 2025 — a 75% quarter-on-quarter increase — with members spending roughly 40% more after joining. Ad revenue grew over 70% year-on-year in Q3 2025, with the ad take rate up over 80 basis points. This is the playbook of a platform that has stopped buying growth and started monetising it.

Brazil is the other story. Shopee’s Brazil GMV reached approximately 60 billion BRL in 2024 — twice Amazon Brazil’s, roughly 40% of Mercado Libre Brazil’s. In Q2 2025, Shopee outgrew Mercado Libre in Brazil 25% to 18%. Mercado Libre responded in mid-2025 by dropping its free shipping threshold to R$19, explicitly targeting Shopee and Temu in low-ticket categories. Average buyer waiting times in Brazil dropped by 1.5 days in Q4 2025 alone as Sea added 300 local Brazilian brands to Shopee Mall.

The risk profile is real. Shopee’s Q4 2025 cost of revenue grew 43.2% — faster than the 35.8% revenue growth — driven by logistics costs as order volumes rose. Adjusted EBITDA grew only 33% in the quarter, suggesting limited operating leverage even as Sea pushed seller take rates higher. Investors took notice.

The structural question, though, is what happens when agents start shopping alongside humans. The standard read is that AI agents threaten marketplace economics. The more accurate read is that agents bifurcate demand into two pipes that both flow through the same infrastructure. Pipe one: human demand — browsing, watching live commerce, buying on impulse, responding to creator content. Live commerce makes this pipe richer in ASEAN, not poorer. Pipe two: agent demand — either humans dispatching agents for specific tasks, or autonomous agents acting on standing instructions. Different UX, different decision logic, different funnel. Both pipes converge on the same fulfilment, payments and trust infrastructure Shopee has spent a decade building.

Agents don’t browse. They consider a shortlist of 50 SKUs, maybe fewer, and decision criteria collapse to price, availability, reviews, fulfilment speed and shelf position. Brand storytelling, lifestyle imagery and video creative are invisible to the agent layer. The product that doesn’t make the shortlist doesn’t exist for agent demand. Which means advertising bifurcates too. Advertising for humans stays emotional, video-led, creator-driven — the Shopee Live world. Advertising for agents becomes payment for preferential shelf placement, pure visibility within the considered shortlist. A portion of brand budget currently sitting on Meta and Google works for humans but doesn’t work for agents. That budget reroutes — and the marketplace shelf is where it lands. Shopee captures budget it doesn’t see today.

Why TikTok Shop Works

Monee: from wallet to digital bank

The biggest brand change of 2025 was the rebrand of SeaMoney to Monee on 8 May 2025, alongside the opening of a 200,000-square-foot headquarters at Rochester Commons in Singapore. Monee now houses two licensed digital banks — MariBank in Singapore and SeaBank in Indonesia and the Philippines — alongside payments, lending, and the MoneeInsure insurance arm. It serves Indonesia, Malaysia, the Philippines, Singapore, Thailand, Taiwan, Vietnam, and Brazil.

The numbers are extraordinary. Monee revenue was $3.8 billion in 2025, up 60% year-on-year. The loan book hit $9.2 billion principal outstanding by end of 2025, up 80%, with $3.2 billion on-book and roughly $1 billion off-book. Non-performing loans past 90 days held at 1.1% — basically the same low default rate as 2024 even as the book nearly doubled. Monee added more than 20 million unique first-time borrowers across 2025.

This is no longer a payment wallet bolted onto an e-commerce platform. It’s one of Southeast Asia’s largest consumer lenders. The rebrand was a signal: SeaMoney was a feature; Monee is a financial institution.

The strategic logic is straightforward. Shopee generates the demand and the data. Garena generates the engagement. Monee provides the rails — and increasingly, the credit. Shopee’s SPayLater BNPL product converts active users into borrowers, and the data flywheel from purchase history feeds underwriting that traditional banks in the region cannot replicate. The Monee growth rate is now meaningfully faster than Shopee’s GMV growth rate, which is exactly what Sea management said would happen.

The 2026 risk is provisions. Q4 2025 saw Monee’s provisions for credit losses up 67%, which dragged adjusted EBITDA growth in the segment to 24.7% — slower than the revenue growth. Sea framed this as confident expansion into new borrower types using AI-assisted underwriting. The market took it as a yellow flag on credit quality. If Monee’s NPL ratio holds through 2026, the bull thesis remains intact. If it deteriorates, Sea has just lent $9 billion across emerging markets in a tightening credit environment.

How Fintech Ecosystems Scale in Asia

What does the Google AI partnership actually mean?

On 19 February 2026, Sea and Google announced an expanded MOU spanning all three Sea businesses. This is the single most important strategic announcement of Sea’s year. Most coverage has framed it as “AI integration to optimise game development and enhance live-ops efficiency.” That undersells it badly.

The MOU has three pillars:

Agentic commerce on Shopee. Sea and Google are co-developing a prototype where an AI agent autonomously browses, compares, and transacts across both Shopee and Google. Imagine telling an agent “find me wireless headphones under SGD 150 with next-day delivery” and having it complete the purchase. Shopee’s 52% market share gives Google immediate scale to deploy and stress-test agentic commerce across 675 million-plus addressable users.

AI-enhanced gaming on Garena. AI-assisted content moderation, personalised in-game support, generative asset production, and developer tooling that compresses development cycles. The strategic objective is defending Free Fire’s live-ops cadence as Tencent, NetEase, and Korean publishers escalate.

Agentic payments on Monee. This is the most structurally novel piece. Monee is participating in Google’s Agent Payments Protocol (AP2), a shared framework for AI agents to initiate and execute payments across platforms. Monee provides Southeast Asia regulatory and product feedback; Google provides protocol architecture. If AP2 becomes the standard rail for agentic payments, Monee is positioned as a co-architect rather than a downstream participant.

The bigger context is the AI cost curve, and this is where the conventional wisdom is wrong. Most investors model AI as a margin headwind for consumer internet platforms. For Sea, it’s a margin tailwind dressed as a headwind. Singapore is the regional AI anchor — SEA-LION (the regional foundation model), SEA-HELM (the multilingual benchmark), and Project SEALD with Google Research are all Singapore-anchored, sitting effectively in Sea’s backyard. Serving Indonesian, Vietnamese, Thai and Filipino content via Qwen-tier and SEA-LION-derived models in production is materially cheaper than running everything on OpenAI or Anthropic. Same cost curve enables more localisation depth, more dynamic live ops, more creator-tool sophistication.

Add in the competitive context. Alibaba released a competing AI model the same week as the Sea-Google MOU, aimed explicitly at the “agentic AI era.” TikTok Shop is investing aggressively in AI-driven creator commerce. The MOU also landed one week after Singapore’s Budget 2026 elevated AI to core national strategy with a Prime Minister-chaired National AI Council. Sea-Google is the regional incumbent’s bet that integration with the world’s leading AI infrastructure beats trying to build frontier capabilities in-house — and that the regional model floor underneath, where the cheap inference actually runs, is the structural margin story nobody is modelling.

For DIA readers, the headline is this: agentic commerce, agentic gaming, and agentic payments are no longer 2030 concepts. Sea is building the prototypes now, with the world’s most advanced AI infrastructure, in the world’s fastest-growing digital market, on a regional model layer that runs cheaper than the Western frontier.

Where does this leave Sea Group in 2026?

Three things matter for the rest of the year.

Free Fire India. A full relaunch in 2026 reopens a market with 8.45 billion mobile game downloads in FY2024-25 (Sensor Tower). Garena’s 2026 bookings guidance — double-digit growth, per management — quietly assumes some India contribution. If that lands, Garena bookings could push past $3.3 billion.

Shopee margin defence in Brazil and Vietnam. TikTok Shop in Vietnam and Mercado Libre in Brazil are both running aggressive shipping subsidy plays. Shopee’s response — VIP membership, ad take rate, SPX logistics — needs to hold the margin line. Q1 2026 results on 19 May will be the first real read.

Monee credit quality. A $9.2 billion loan book growing 80% per year in emerging markets with rising provisions is either the great fintech compounding story of the decade or a setup for a write-down. The 1.1% NPL ratio is the number to watch.

What’s emerging is a company that increasingly looks like the dominant context-owner of an entire region. Three flywheels feeding each other, with persistent customer state — gaming graph, payments, purchase history — as the underlying compounding asset. The Google partnership is the bet that the next layer of value capture happens at the AI-agent layer, not the user-interface layer. If they’re right, Sea owns the rails for an entire generation of Southeast Asian digital commerce. If they’re wrong, they’ve handed the AI advantage to a hyperscaler that doesn’t depend on them.

The market still prices Sea as three businesses bolted together. The reality is closer to one playbook run three times — gaming live ops, commerce live ops, fintech live ops — on top of a regional AI stack that compresses cost faster than peers can match. Gametime is the new primetime, in markets the West is still learning to read.

This is the most important digital stack in the region. 2026 is the year we find out how strong it really is.