Software, AI products and semiconductor design can generate foreign exchange with far less material throughput than heavy industry. The FY2026-27 budget has the ambition; it now needs a complete product-and-talent ecosystem.
For Bangladesh, one additional taka invested in digital productive capacity can often travel further than the same taka spent in a conventional, import-intensive industry. This is not a universal accounting identity: a failed app can destroy capital as efficiently as a failed factory. But software and digitally delivered services combine high value added, low freight cost, rapid replicability and modest physical inputs. Once a product is built, the next international customer can be served at a marginal cost no containerised good can match.
The FY2026-27 budget's technology ambition economically important. It places the ICT sector's current GDP contribution at only 1-2% and targets 10% within five years. It proposes 5G coverage for 90% of the population, 100 Mbps broadband, a National Fibre Bank, a One Citizen-One ID-One Digital Wallet architecture, specialised laboratories for AI, semiconductors, robotics, machine learning and big data, and a Tk 500 crore start-up fund. It also targets 200,000 new technology jobs a year and 800,000 indirect opportunities through freelancing and creative work.
Digital exports are attractive because they do not require imported cotton, fuel-intensive furnaces, chemical processing, refrigerated logistics or physical inventory for every sale. A software subscription can be exported repeatedly through existing networks and paid for in foreign currency. Compared with many manufacturing activities, the direct energy and water used per dollar of value added can be much lower, especially for software, design, business-process and engineering services.
ICT is not environmentally neutral. Data centres, AI training and semiconductor fabrication can be extremely electricity- and water-intensive. Bangladesh should therefore avoid the slogan that all digital activity is green. The strategic distinction is between asset-light software and design on one side, and hyperscale compute or wafer fabrication on the other. The country can first specialise in AI applications, cloud software, embedded systems and chip design, capturing intellectual value without immediately assuming the capital and resource burden of an advanced fabrication plant.
Bangladesh should stop measuring digital ambition by the number of
trainees and start measuring it by products, recurring revenue, patents,
retained customers and foreign subscriptions.
The global opportunity is unusually large. UNCTAD projects that the AI market could reach US$4.8 trillion by 2033. Global semiconductor sales, having reached about US$792 billion in 2025, are now on track to surpass US$1 trillion in 2026, with the industry's latest 2026 forecast pointing toward roughly US$1.5 trillion as AI infrastructure demand accelerates. Semiconductor design is the most realistic entry point for Bangladesh, verification, physical design, embedded firmware, analogue layout and application-specific intellectual property can be supplied across borders by skilled teams.
Bangladesh should stop measuring digital ambition by the number of trainees and start measuring it by products, recurring revenue, patents, retained customers and foreign subscriptions.
Bangladeshi professionals and firms are rapidly adopting American and other foreign AI tools. International-card statistics show rising spending abroad across education, professional and digital categories including cloud, software, design and AI subscriptions far more readily than the world pays for Bangladeshi subscription products.
There is still no Bangladesh-origin AI product with globally significant recurring usage. That indicts the ecosystem, not local talent. Freelancing rewards hours sold; a product economy rewards intellectual property sold repeatedly. A freelancer serves one client at a time, whereas a software-as-a-service and AI-as-a-service company can serve 10,000 customers with its core code written and owned in Bangladesh.
Previous administrations advertised large digital numbers, infrastructure programmes and training counts that were often difficult to reconcile with export receipts, company growth and product outcomes. Public allegations of politically connected procurement and weak value-for-money require independent audit, not the partisan replacement of one set of claims with another. The government should publish project-level expenditure, beneficiaries, source-code ownership, utilisation, procurement competition and export results for every major ICT programme.
The budget has components, but not yet a complete production system. Connectivity without affordable cloud compute is incomplete. Training without commercial mentors produces certificates. A start-up fund without procurement access produces pitch decks. Semiconductor laboratories without electronic-design-automation licences and industry tape-out partnerships produce demonstrations rather than chips.
The first priority should be a National AI Product Fund, not another generic start-up pool. Matching grants should support Bangla and regional-language AI and applications in health, agriculture, trade compliance, finance and industry. Government should become a disciplined first buyer through challenge-based procurement with strict cybersecurity, privacy and performance tests, and exportable intellectual property should remain in Bangladesh.
Second, establish a fabless semiconductor mission. India is training 85,000 engineers through its Chips to Startup programme, while Vietnam has attracted chip-testing and AI research centres. Bangladesh should create university design centres linked to EDA vendors, provide shared licences and cloud access, fund 20-30 prototype tape-outs a year, and recruit diaspora designers. The goal is not a leading-edge fab; it is a credible design-services and IP pipeline.
Third, connect ICT to physical industry. AI-powered light engineering can improve computer-aided design, machine vision, predictive maintenance, energy optimisation, quality inspection and small-batch manufacturing. Exportable software should emerge from solving domestic factory problems. This is the home-market mechanism in development economics: a demanding local market creates the learning, scale and reference customers that support international competitiveness.
Fourth, repair the commercial plumbing. Start-ups need employee stock options, convertible instruments, bankruptcy rules for honest failure, fast foreign-currency collection, predictable cloud taxation, data protection, patent support and one-window export documentation. Universities should license IP and allow researchers to form companies. Banks need revenue-based and intangible-asset lending tools.
Bangladesh will not become an AI economy by buying more foreign AI. It will do so by converting local problems into global products. The FY27 budget's targets are bold enough to mobilise attention, but targets do not compound; products do. A country that can export garments to the world can also export algorithms, chip designs and industrial software. The next great Bangladeshi factory may have no chimney, no warehouse and no container yard. Its raw material will be talent; its machinery will be compute; and its foreign-exchange earnings will arrive one subscription at a time.
The writer is a Trade and Development Economist, working as the Chief Research Officer (CRO) of Textile Today Innovation Hub and Co-Founder & CEO of Nexadev, an AI-powered solution development firm
