Petro Matad, the AIM-listed Mongolian oil producer, saw its shares fall on Thursday after releasing an operational update that failed to spark investor enthusiasm.

The company said the long-awaited 2026 Oil Sales Agreement with PetroChina is now close to being signed off, clearing the way for a year’s worth of stored crude to finally hit the market.

With roughly 35,000 barrels sitting in Block XIX storage tanks and oil prices climbing throughout the year, the company has the chance to sell at a substantially better price than if the crude had been sold month by month.

Approval stalled over issues involving the refinery contract, but Petro Matad now says that the issue is resolved and sign-off should follow shortly. As a backstop, it has also been sounding out independent Chinese oil traders, though the established export route remains its clear preference.

In terms of production, Heron-1 averaged 126 barrels per day over the first quarter with a low, stable water cut of around 3%, while Gazelle-1 has settled at roughly 123 barrels per day, comfortably ahead of forecast after water breakthrough was first detected. These are steady numbers, but hardly anything to get excited about.

Encouragingly, operating costs haven’t crept up since Gazelle came on stream last November, thanks to power savings from connecting Heron-1 to the national grid. The company is now weighing whether to hook up Gazelle too. Block XX has now produced more than 110,000 barrels since start-up.

Where there may be investor frustration is developments in a farm-out agreement, or a lack thereof. Talks with the most advanced potential farm-in partner are dragging on, with the partner opting to conduct further evaluations. Petro Matad management said it was ‘disappointed’ at the lack of progress.

Investors were evidently disappointed as well, and shares were 9% lower in early trading on Thursday.

Computacenter has struck a deal to acquire Government Acquisitions Inc (GAI), a US value-added reseller focused on the federal government sector, in a move that opens up a significant new growth avenue for the technology and services group.

The deal, worth an enterprise value of up to $92m, has already been cleared by the US foreign investment regulator CFIUS and is expected to be completed on 1 June 2026.

Computacenter will make an initial cash payment of $63m, with further performance-based payments of up to $29m, payable through the end of 2027. The acquisition is being funded from existing cash and is expected to be immediately earnings accretive.

Cincinnati-based GAI gives Computacenter a foothold in the US federal government market, a sector that has been difficult to break into.

Mike Norris, CEO of Computacenter, said: “We are proud of the relationships we have built with public sector customers across Europe and Canada and are delighted to have the opportunity to bring one of the leading US federal government VARs into Computacenter. GAI provides us with access to a new market for growth in the United States, diversifies our business and leverages our growing capabilities and infrastructure.”

The business brings over 35 years of experience, around 90 staff and deep relationships across federal agencies. It was named Nvidia’s US Public Sector Partner of the Year for 2025, a notable credential given the surging demand for AI infrastructure across government.

In 2025 GAI reported gross invoiced income of roughly $390m and adjusted EBITDA of about $8m.

Analysis for informational purposes only. Capital at risk.

The Memory Ceiling in the AI Industry: Google’s decision to cap Gemini tokens signals an AI hardware bottleneck, from HBM to server DRAM. The era of unmetered compute access is ending.

The Ecosystem Toll: By embedding agentic AI across Search, Gmail, and YouTube, Google transforms its product suite into an ecosystem. A “base quota + pay-as-you-go” model protects margins on the downside and captures agent-driven token volume on the upside.

The Super App Readiness: China’s WeChat, Alibaba, and ByteDance already operate the integrated environments Google is racing to build. For investors, the critical question is which ecosystem model captures the agentic era’s economics.

Google latest annual product showcase (I/O 2026) wasn’t just a regular product rollout.

It signals a structural shift in the AI industry: the era of unmetered compute is ending, and ecosystem integration is the new monetisation lever.

The “Memory Ceiling” in the AI Industry

Google put token usage caps on Gemini subscribers. The move signals a supply chain constraint in the AI industry, not a product strategy.

Global memory shortage: High-Bandwidth Memory (HBM) and server DRAM contract prices have risen sharply. Samsung’s HBM capacity is sold out through end-2026. SK Hynix signals demand will outpace supply for three years — new factory capacity does not come online until late 2027.

Alphabet’s cloud contract backlog increased by 5x within one year. Its CEO admitted cloud revenue would have been higher if the company could meet demand.

Chinese solution: Chinese models such as DeepSeek have optimised for compute efficiency using Mixture-of-Experts architecture, model distillation, and quantisation. The result is two distinct AI markets: Western hyperscalers optimise for raw throughput, while Chinese platforms maximise yield per compute unit.

Memory supply for China: ChangXin Memory Technologies (CXMT), China’s leading DRAM maker, is preparing to list in Shanghai in 2H26. It currently produces consumer DDR4 and DDR5 but targets HBM production by end-2026, potentially alleviating China’s memory supply constraint in coming years.

China’s architecture efficiency gains and its memory build-out address different parts of the AI supply chain.

Usage-Based Pricing Captures the Token Economy

Google moved subscriptions from flat-rate access to “base quota + pay-as-you-go.”

Subscribers exceeding their quota are throttled to a lighter model or charged for additional tokens. Enterprise cloud has operated on consumption-based pricing for a decade. Porting this model to consumers is a logical next step.

Downside protection, volume capture: The base quota protects against margin pressure from uncapped compute exposure. Pay-as-you-go captures the volume upside as agentic workflows push token consumption 10-50x higher per query.

The token economy potential: The total daily output tokens generated by all AI models is projected to exceed the combined spoken and written output of the world’s population by June 2026. Google Cloud’s first-party models alone now process 23 trillion tokens daily, up from 14.4 trillion last quarter. OpenRouter’s weekly token volume rose by 6x YTD.

The fastest-growing use case is agentic inference: multi-step workflows where models plan, retrieve context, revise, and iterate. A simple chatbot requires around 30,000 tokens for a summary task; an AI agent executing a comparable programming task can consume up to 20 million tokens.

Tokens consumption as a corporate KPI: Token consumption is becoming an internal performance metric at the largest tech companies.

Meta’s 85,000 employees burned through 60 trillion tokens in 30 days via an internal AI leaderboard called “Claudeonomics”. Meta’s CTO said one top engineer spends the equivalent of his salary on tokens and reportedly 10x’d his output. Amazon and Microsoft have followed suit, with employee performance reviews now incorporating AI usage impact metrics.

The token demand is so high that budgets are breaking. Uber exhausted its entire 2026 AI budget by April after Claude Code adoption jumped from 32% to 84% of its engineers. Monthly API costs per engineer ran $500 to $2,000. By spring, 70% of Uber’s committed code originated from AI tools.

China’s spot market assault: Chinese LLMs captured about 40% of OpenRouter token volume currently, up from 10% in January, while the combined share of US hyperscalers on the same platform dropped from 70% to 40% over the same period.

Aggregators attract cost-sensitive developers and SMEs who use dynamic routing: routine tasks go to the cheapest capable model, while complex queries are reserved for premium Western models. Chinese LLMs have captured the bulk of this routine compute by delivering 80–90% of Western flagship performance at roughly 20% of the cost.

The Agentic Bundle: Dismantling the App Layer

Google integrates agentic capabilities across its entire product line, translating its product portfolio into an ecosystem.

Multiple features, one strategy:

• The AI Search Box lets users find and book hotels natively, bundling items into a unified “Universal Cart” that bypasses individual merchant apps and carries a potential take-rate for every transaction.
• Gmail Live turns the inbox into a voice assistant.
• Ask YouTube answers technical questions by jumping to the exact frame containing the solution.
• Daily Brief aggregates Workspace data overnight into a personalised morning summary.

Dismantling the app layer: By offering a one-stop, seamless agentic ecosystem to users, Google essentially disintermediates individual apps and captures the corresponding monetisation opportunities.

The Super App Advantage: China’s Head Start

Google’s products are currently fragmented: browser (Chrome), search, email (Gmail), video (YouTube), documents (Docs), cloud (Drive), payments (Google Pay), messaging (Messages), and meetings (Meet). Each requires a separate interface and workflow.

Google is trying to integrate these silos through agentic AI — a single assistant that moves across them on the user’s behalf.

In China, the super app already is the OS.

Tencent: WeChat handles communication, payments, commerce, document management, ride-hailing, travel booking, and financial services — all without the user ever leaving the app.

Alibaba: Alibaba’s ecosystem runs e-commerce, cloud infrastructure, enterprise workflows, and payments under one roof.

ByteDance: Embeds AI directly into content, commerce, and creation inside Douyin. The user does not switch contexts. The context lives inside the app.

Baidu — the exception: Baidu has a search engine comparable to Google Search, but lacks the commerce, communication, and enterprise workflow breadth of its peers. Without the ecosystem, its Ernie Bot has struggled for traction.

The App-Only Trap

The agentic era creates a structural problem for standalone apps. A standalone app cannot offer agentic workflows spanning email, search, calendar, and payments. A pure-play AI chatbot cannot execute transactions across disconnected platforms.

The structural threat: Take Booking.com as an example. Its business model takes a 15-20% commission on every hotel booking and flight. Google’s Universal Cart lets users search, compare, and book hotels directly from the AI Search Box, bypassing Booking’s entire intermediation layer.

A three-way dilemma: Integrate with Google’s cart and give up margin. Build its own AI travel agent and burn capital against Google. Or do nothing and watch transactions migrate to Google’s checkout.

Meituan — the China example: Meituan has deep local commerce capabilities but its major competitor, Alibaba, has integrated local commerce businesses (Ele.me) into the agentic workflow, potentially threatening Meituan’s market position.

The next phase of the AI race is shifting from model quality to workflow ownership. Google has signalled the pricing model for that race. China’s super apps have already built the roads.

This article is a “periodical publication” for information only and is not investment advice or a solicitation to buy or sell securities. This article does not constitute a “personal recommendation” or “investment advice” under UK FCA regulations. Investing in equities involves significant risk. The author holds NO position in the securities mentioned. There is no warranty as to completeness or correctness. Please do your own due diligence or consult a licensed financial adviser. Please read the Full Disclaimer before acting on any information. Images created with the assistance of AI.

Article provided by Asia Pulse.

Specialist coatings services provider Hardide (LON: HDD) has won a significant order worth £2.4m in the energy sector in North America. The US facility has improved its operational efficiency. This has increased the earnings forecast by 17% to 4p/share. The share price is 14.7% higher at 58.5p, which is less than 15 times prospective earnings.

Sovereign Metals (LON: SVML) says monazite containing heavy rare earths have been confirmed in multiple pits at Kasiya in Malawi. These include Dysprosium (Dy), Terbium (Tb) and Yttrium. These can be recoverable from tailings. The share price gained 10.5% to 37p.

Defence equipment and services provider Cohort (LON: CHRT) says full year trading is ahead of expectations, and the order book has reached the record level of £620m and £253m of the order book covers 80% of 2026-27 expectations. Operating profit will improve from £27.5m to £36m. The growth is coming from the communications and intelligence division. The share price increased 8.35% to $13.37.

Mathematical modelling and biostatistics company Physiomics (LON: PYC) has secured new contracts worth a total of more than £345,000 so far this month. Some of the projects will start imminently and they will produce revenues until the end of 2027. The share price rose 7.69% to 0.7p.

FALLERS

Shares in Sound Energy (LON: SOU) have fallen a further 21.8% to 2.15p following yesterday’s news that it is selling its development assets in Morocco for $57m in cash and relinquishing nearby exploration assets. The share price is at its all time low. Ten years ago, Sound Energy was one of the larger companies on AIM and it is currently worth £6.3m.

Health and safety consultancy PHSC (LON: PHSC) shares have declined an additional 8.82% to 7.75p following yesterday’s trading statement previewing a full year loss. There was an improvement in the second half. The annual results will be published in July.

Staffing provider RTC (LON: RTC) says that the trajectory of positive trading in the first quarter of 2026 and six major contracts have been won and rail maintenance demand is in line with 2025 levels. Rising costs will hit margins in rail and energy divisions, and it is also holding back activity levels in the second quarter. Permanent recruitment vacancy levels are at their lowest point since 2021. Daavid Stredder has put forward AGM resolutions for the appointment of Paul Hooper, former Alumasc boss, as independent chairman and Gerard Oates as an independent non-executive director. He objects to Andy Pendlebury being chairman and chief executive and there being only one independent non-executive director and he is employed by the company’s broker. He also complains about the rise in board pay. David Stredder also opposes the appointment of new director Andrew Kitchingham because of the number of directorships he currently has. The share price slipped 8.16% to 112.5p.

Defence contractor RC Fornax (LON: RCFX) interim revenues dipped from £2.5m to £2.2m and it fell into loss. There is improved revenues visibility Cavendish has maintained its 2025-26 forecast loss at £2m. Order momentum is improving. The share price fell 7.93% to 7.55p.

Union Jack Oil (LON: UJO) has published further information about AGM resolutions 7 and 10. Non-executive director John Americanos, who is the largest shareholder with 6.78%, disagrees with the other directors on the resolutions. Resolution seven is for the appointment of Craig Howie as a director. The board, other than John Americanos, is against Craig Howie being reappointed to the board following his removal as an executive director in January, prior to the appointment of John Americanos. The argument was that he did not have the knowledge for his job and breaches of confidentiality. A nominee shareholder requested that resolution 7 was included at the AGM, as well as resolutions for the re-election of directors. Resolution 10 would reduce the nominal value of shares from 5p, which is above the market price, to 0.05p, so that shares can be issued. The share price declined 7.5% to 3.7p.

Cohort has delivered another year of solid growth, with both revenue and profit coming in ahead of City expectations, as its defence and security technology businesses continued to win new business amid rising global defence spending.

The group reported revenue of £303m for the year to 30 April, up 12%, with adjusted operating profit of around £36m. Net margin improved to 11.9% from 10.2%.

Cohort shares rose 13% on the news.

The standout was the Communications and Intelligence division, where revenue jumped to £159m from £125.4m and margins pushed to roughly 20%.

A full-year contribution from EM Solutions drove much of the improvement. Sensors and Effectors was steadier, with revenue broadly flat at £144m and a thinner 7% margin, reflecting cautious trading on ELAC’s Italian sonar contract and the sale of SEA’s high-margin transport business last June.

All three businesses in the division traded profitably, though the board acknowledges the margin remains below its mid-term target.

Investors should be encouraged by the order book. Intake of around £313m boosted the closing order book to a record £620m, with work now stretching out to the mid-2030s. EID, MASS and EM Solutions led the charge, alongside Chess and SEA, helped by wins such as a €42.3m contract for EID to supply integrated communication systems to the Portuguese Navy.

The order book already underpins around £253m of revenue for the new financial year, or roughly 80% cover.

Andrew Thomis, Chief Executive of Cohort, said: “Cohort performed strongly in 2025/26, exceeding market expectations. Following another year of strong order intake, our closing order book has reached a new record level, and we have encouraging prospects for further orders.

“I’m particularly pleased with the strong maiden full year contribution from EM Solutions. The business has made good progress in capturing opportunities for growth and collaboration with other Cohort subsidiaries.”

Physiomics shares rose on Wednesday after it announced it has secured multiple new and follow-on contracts worth more than £345,000 in total.

The awards come shortly after a shakeup of the board that investors will hope breathes fresh life into the firm.

The AIM-listed company, which supports drug developers with mathematical modelling, biostatistics and data science, said the awards secured so far this month span a mix of new and follow-on work with UK and international clients. Among them are a NASDAQ-listed clinical-stage biotech, several clinical-stage therapeutics developers and a number of leading cancer research groups.

The projects will draw on Physiomics’ expertise across modelling and simulation, biometrics and data science to support key stages of drug development, including Phase 1 and 2 clinical data analysis, first-in-human dose selection and ongoing clinical programme support.

Encouragingly for investors, several are expected to start imminently and will deliver revenue through to the end of 2027, giving the company both near-term income and useful medium-term visibility.

The board reckons the awards also underline the deepening quality of its client relationships, with repeat business pointing to a strengthening reputation as a trusted scientific partner across the biotech and pharma sectors.

Dr Peter Sargent, CEO of Physiomics, commented: “We are delighted to have secured these contract awards. They reflect our continued investment in business development and marketing, as well as the team’s success in expanding and diversifying our pipeline. These opportunities further strengthen our order book and improve revenue visibility across the business. Our broader pipeline remains strong and we’re excited about a number of additional opportunities that we hope to be able to announce soon”.

Pets at Home has slashed its dividend by more than 40% after a difficult year in which both revenues and profits fell, though the retailer pointed to early signs of recovery and the continued strength of its Vet business.

For the 52 weeks to 26 March 2026, statutory revenue fell 0.8% to £1.47bn, while group underlying pre-tax profit dropped 30.2% to £92.8m.

The dividend was cut to 7.4p from 13p the year before, a 43% reduction, and is now set at 50% of earnings under a refreshed capital allocation policy announced at the pre-close update.

Underlying earnings per share fell 29.7% to 14.8p.

Pets at Home’s issues have been reflected in its share price, which is down by more than 60% since its 2022 peak.

The bulk of the pain came from retail, where consumer revenue dipped 1% to £1.29bn in challenging market conditions and underlying profit collapsed 57.8% to £30.8m.

Retail free cash flow fell to just £2.7m, down from £30.6m, largely tracking the profit decline. Group gross margin overall fell around 120 basis points to 45.7%.

This isn’t good news for investors, and the dividend cut shouldn’t have come as a surprise.

Management was candid that the dividend cut and the profit reset reflect a Retail business that had drifted off course, prompting the Retail Turnaround Plan, launched in the third quarter, centred on four priorities: product, price, execution, and cost. However, it doesn’t seem to be filtering through to financial performance.

A November price cut on more than 1,000 food products, averaging 12%, was a deliberate margin sacrifice aimed at restoring volume, and there are signs it is working, with food volumes up 3.7% in the fourth quarter.

Despite the setbacks, there were reasons to be encouraged. The Vet Group again outperformed, with consumer revenue up 5% to £688m and underlying profit up 10.4% to £83.8m, helped by strong Care Plan sign-ups.

The business remains the clear leader in UK pet care, with 460 pet care centres, 7.4m active Pets Club members, brand awareness above 95% and well-invested distribution and digital infrastructure now in place.