Grab Bets Big on AI to Triple Profit by 2028 — What That Means for the AI Industry

How Southeast Asia’s superapp plans to turn recommendation engines, agents, and autonomous rails into the kind of profit curve investors dream about

A Grab driver in Jakarta pauses on a delivery, watching a notification that an in‑app assistant has already rebooked their next job and suggested a grocery item a frequent customer will likely add. The scene is small and ordinary, but it captures a collision between two worlds: a mobile workforce still paid by the trip and software that promises to stitch those trips into predictable, higher‑margin flows. That tension is where the real story begins.

Most observers hear the headlines and take the obvious reading: Grab is simply expanding into groceries and fintech to squeeze more revenue from an enormous user base. The subtler, and underreported, implication is that Grab sees AI not as a neat cost cut but as an operating system switch that will change who sells, ships, and underwrites services across Southeast Asia. This matters to AI builders because it converts model access into durable commercial leverage faster than a chatbot or a one‑off integration can. According to Reuters, Grab’s executive team says AI and new services are central to its plan to triple EBITDA to $1.5 billion by 2028. (investing.com)

Why the industry should stop treating Grab like just another ride app

Grab already operates at a density few startups can buy: over 50 million monthly transacting users and billions in annual GMV. That scale makes it a uniquely attractive playground for production AI, from personalized agent assistants for drivers to automated pricing across deliveries. The company’s own results show the leverage: Grab reported its first full‑year net profit in 2025 and explicitly tied a $1.5 billion 2028 Adjusted EBITDA target to increased efficiency and new service lines. (grab.com)

Competitors and the new battleground for AI agents

The obvious competitors are regional superapps and global platforms trying to localize services. But the real fight is for the agent layer and the data plumbing that ties agent outputs to transactions. OpenAI’s recent push to turn ChatGPT into an app platform and build AgentKit makes it cheaper to ship agent workflows at scale, which will accelerate enterprise deployments that companies like Grab can either buy into or replicate. This platformization of agents creates a race between model providers and vertically integrated superapps to own the execution layer where money actually changes hands. (tech.yahoo.com)

The core story: where the math of AI meets unit economics

Grab’s financial position is unusually robust for a company still transitioning from growth to profitability. Management exited 2025 with about $7.4 billion in gross cash liquidity and guided 2026 revenue to roughly $4.04 to $4.10 billion while setting the $1.5 billion Adjusted EBITDA 2028 target. Those numbers create a staging ground for capital‑intensive bets such as autonomous mobility, which Grab is piloting with partners and sensor distribution deals. If AI reduces incentive spend on matching and increases lifetime value per user by just 10 percent, the lift to gross margins compounds across tens of millions of users. (ainvest.com)

If Grab turns its frequent, data‑rich app into a platform of agents that close transactions, it will convert AI experiments into recurring revenue, not just product features.

What this means for AI vendors and model shops

For LLM vendors and systems integrators, Grab’s strategy is an invitation and a warning. The invitation: embed deeply into the transaction path and capture recurring revenue from API calls, custom models, and agent orchestration fees. The warning: if a superapp with Grab’s scale decides to retrain on its own proprietary signals and build bespoke agents, third parties become commoditized in weeks, not years. The Information reported that Grab has been heavy on internal AI usage and that investors are parsing whether the company’s stock reflects these investments, signaling that execution risk is front and center. (theinformation.com)

The cost nobody is calculating for AI at scale

Training and inference are expensive, but the larger bill is governance and integration. Hooking an agent into payments, loan underwriting, and real‑time logistics requires connectors, human review layers, and traceability. If Grab’s agents are used to underwrite microloans or to autonomously dispatch AV shuttles, the company will need to invest in safety, monitoring, fraud controls, and regulatory teams. Those are not one‑time engineering costs but ongoing operating expenses that grow with volume. A sensible finance team will model an incremental $0.02 to $0.10 per transaction in AI governance costs for mid‑complexity agent workflows, which eats into gross margins unless the revenue uplift scales faster. Dry observation: venture boards love growth stories until someone mentions logs and audits, then the tone changes.

Practical scenarios for businesses evaluating Grab’s move

A grocery chain selling through GrabMart could see conversion rates rise from 4 percent to 6 percent if a Grab agent surfacing personalized bundle suggestions increases basket size by 25 percent. For a business making $10 average order value and 100,000 annual orders via the platform, that translates to an incremental $500,000 in gross merchandise value and roughly $35,000 to $70,000 in incremental gross profit, assuming platform fees and marginal cost structures. Multiply that across millions of orders and it becomes material for supermarkets and FMCG brands looking to justify bespoke integrations. The math is simple; the orchestration is not.

Risks and hard questions that will test the claim

Automation timelines are optimistic. Autonomous mobility remains capital intensive and highly regulated, and the revenue upside depends on mass adoption that may take longer than the 2028 window implies. AI safety and data privacy laws across Southeast Asia are uneven, creating compliance drift that could slow rollouts. Lastly, owning both the transaction and the model creates concentration risk; a systemic model failure or a targeted adversarial attack could cascade through payments, lending, and logistics simultaneously. These are solvable problems but they are expensive and time consuming.

The playbook for startups and AI teams that want in

Sell where Grab cannot or will not go deep. That means industry‑specific models, turnkey safety auditing, or niche orchestration layers that stitch into Grab without becoming replaceable. Offer measurable ROI within 90 to 180 days. If an integration can prove it reduces incentive spend or increases repeat purchase rates within a quarter, it will get prioritized. One can be clever and binary about this: if the product can drive 1 to 3 percent lift in user monetization or 5 to 10 percent drop in operating cost per order, it becomes interesting. If not, it becomes a feature.

Where this leaves the AI industry in 2028

If Grab succeeds, the industry will see more vertically owned model stacks built around high‑frequency consumer platforms. That will accelerate demand for specialized inference infrastructure, agent orchestration frameworks, and enterprise safety tooling. If Grab fails, the lesson will be the limits of scale without operational finesse. Either way, the winner will be the team that turns agent outcomes into reliably measurable cash flow.

Final note looking forward

Grab’s plan is a real test of whether AI can be an engine of sustainable margin improvement when woven into commerce, mobility, and finance, not just a flashy lab demo. The industry should pay attention because the outcome will change the shape of enterprise AI budgets and the commercial role of models.

Key Takeaways

• Grab plans to triple Adjusted EBITDA to $1.5 billion by 2028 using AI and new services, turning user frequency into higher margins.
• The company’s 2025 profitability and $7.4 billion liquidity give it unusual runway to fund capital‑intensive AI and autonomous bets.
• Platformization of agents, exemplified by recent developer ecosystems, shifts value to whoever owns execution and data.
• Vendors should focus on measurable, short‑term ROI and safety tooling if they want strategic partnerships rather than one‑off integrations.

Frequently Asked Questions

How will Grab’s AI bets affect pricing for AI services in Southeast Asia?
Lower marginal pricing for generic API calls is likely as demand scales, but custom models trained on Grab’s proprietary data will command a premium. Enterprises should expect volume discounts for raw inference but higher fees for integrated agent orchestration and compliance services.

Can small AI providers still work with Grab without being acquired?
Yes, by offering vertically specific capabilities that are costly to develop in‑house, such as domain tuned models, regulatory compliance stacks, or proven fraud detection pipelines. Prove a clear ROI and short payback period to avoid becoming a short‑term vendor.

Will Grab’s agents replace drivers and merchant staff quickly?
Automation will reconfigure roles over time; agents can optimize dispatch and automate routine tasks, but full replacement, especially in logistics and onsite retail, will take years and regulatory approvals. Upskilling and new job categories are more likely in the medium term.

Is the 2028 profit target realistic for the AI industry to support?
It is plausible if AI materially cuts matching and incentive costs and new high‑margin services scale as planned. Execution risk, regulation, and integration costs make the target ambitious but within the realm of possibility given Grab’s liquidity.

How should AI teams prioritize features to win Grab’s business?
Prioritize features that move payment or underwriting behavior directly: cart personalization that increases average order value, real‑time fraud prevention, and lending models that reduce default rates are high impact and measurable quickly.

Related Coverage

Explore how autonomous shuttles are reshaping urban mobility and what sensor supply deals mean for regional startups. Also read about AI agent platforms and the economics of model provisioning to understand the vendor side of this race. The AI Era News has ongoing reporting on agent toolkits and platform strategies worth bookmarking.

SOURCES: https://www.investing.com/news/stock-market-news/singapores-grab-bets-on-ai-new-services-to-triple-profit-by-2028-4526620, https://www.grab.com/sg/press/others/grab-reports-fourth-quarter-and-2025-results-with-first-full-year-net-profit/, https://www.ainvest.com/news/grab-2025-breakthrough-strategic-pivot-autonomous-mobility-infrastructure-2602/, https://www.theinformation.com/briefings/singapores-grab-reports-first-full-year-profit, https://tech.yahoo.com/ai/chatgpt/articles/openai-just-turned-chatgpt-app-204723722.html