[Test] This week: IFRS 18: the 2026 comparative needs work now, and 2 more
The Finance Decoder
Finance teams tend to trust the numbers that tie out. This week's items look at where that trust gets tested: budgets, departmental plans and a model that passed every check.
IFRS 18 transition: restating the comparative year, step by step
For a December year-end group, the 2026 comparative has to be restated under IFRS 18, and it first appears in the Q1 2027 interim drafted in April 2027, not in the annual report. With under three months of 2026 left, the order of work matters.
Appendix C applies the standard retrospectively under IAS 8 (C2). In place of the usual line-by-line adjustment, C3 requires a reconciliation for each line item in profit or loss between the restated comparative and the amounts previously presented under IAS 1. The work does not shrink. It moves into one table that the auditor will tie back to statements already signed.
For quarterly reporters, C4 requires the first year's interims to present every IFRS 18 heading and the subtotals in paragraphs 69 to 74. C5 adds a reconciliation for the comparative periods. So the restated comparative has to exist quarterly: Q1 2026, then Q2 and H1 2026, then Q3 and nine months, then the full year.
Start from the 2026 trial balance, not the published income statement. Published lines are aggregates, and finance income and other income are the usual casualties. Map at account level, then at transaction level where an account mixes items. In the article's illustration, rental income on investment property is investing, interest on lease liabilities is financing, and foreign exchange on trade payables is operating. Profit for the year stays at 216 under both standards.
Management-defined performance measures get no Appendix C relief. On my reading of paragraphs 31, 32 and 123 together, each measure reported in 2027 needs a 2026 comparative reconciled to the IFRS 18 subtotal.
The plan front-loads 2026. Finish the main business activity assessment in October. Map the chart of accounts in November and split mixed accounts at source. A mixed account is far cheaper to split before year end than to analyse transaction by transaction in March.
IAS 23: when an off-plan development must capitalise its borrowing costs
Whether an off-plan developer capitalises its borrowing costs depends on the revenue conclusion. The loan has nothing to do with it. Two developers can borrow the same amount at the same rate for the same tower. One carries the interest in inventory for three years and releases it through cost of sales at handover. The other charges it to finance cost as it accrues. Both can be right.
IAS 23.8 requires capitalisation of borrowing costs directly attributable to a qualifying asset. There is no policy choice. The catch is that the developer may never recognise the tower as an asset. If control of each unit passes at handover, the developer holds inventory under construction, and the interest on it is capitalised. If control passes over time under IFRS 15.35(c), the developer holds a receivable or a contract asset instead. The IFRS Interpretations Committee's March 2019 agenda decision concluded that neither is a qualifying asset. The interest then goes to profit or loss as incurred, and gross margin rises by the same amount.
Measurement has two routes. For a specific facility, IAS 23.12 allows the actual cost less investment income on idle funds. For general borrowings, IAS 23.14 applies a weighted average capitalisation rate to the spend, and the amount capitalised cannot exceed the interest actually incurred.
Timing matters as well. IAS 23.20 suspends capitalisation during extended periods when active development is suspended. IAS 23.22 stops it when substantially all the activities to prepare the asset are complete.
In the article's example, one year of construction capitalises AED 29.0 million and expenses 42.0 million. If every unit had been sold over time, the full 71.0 million would be finance cost.
The file needs three papers: the IFRS 15 conclusion for each contract type, the funding map, and the dates. Get the first wrong and every number downstream moves.
Service charges under IFRS 15: principal or agent?
A landlord that recharges cleaning, security and utilities to tenants can show the recovery as gross revenue only where it is the principal for that service. Where it is an agent, revenue is the fee it keeps, often nil. Profit is the same either way. The revenue line, the margins and any covenant measured on revenue are not.
Services in a lease are non-lease components. IFRS 16.17 sends the lessor's allocation to IFRS 15, so the income is revenue. Some recharges are not services at all. IFRS 16.B33 covers costs that transfer nothing to the lessee, such as insurance on the landlord's own building.
The principal test sits in IFRS 15 paragraphs B34 to B38. Identify the specified service, then ask whether the landlord controls it before it reaches the tenant (B34A). Paragraph B35A includes the right to direct another party to perform the service on the entity's behalf. Paragraph B37 gives three indicators: primary responsibility, inventory risk and pricing discretion. Vacancy works like inventory risk. If two floors are empty under a fixed cleaning contract, the landlord bears that cost.
Cleaning, security and common area maintenance under contracts the landlord manages usually point to principal. Metered tenant electricity billed at the provider's tariff often points to agent. An investor recharging a building's service charge has not contracted with the cleaner and cannot direct them. Under Dubai Law No. 6 of 2019, common parts are managed by the developer or a management company selected by RERA. That points to agent.
In the worked example, three contracts cost AED 3,000,000. At 80 per cent occupancy the landlord recovers 2,400,000 plus a 240,000 fee. Gross revenue is AED 2,640,000, and the 360,000 shortfall shows as a cost.
Keep a short memo for each recharged service. Paragraph 123 of IFRS 15 requires disclosure of the judgements that significantly affect revenue, and on a large portfolio this one usually qualifies.
Keep the budget fixed and let the forecast move
Corporate Finance AI argues the approved budget should stay as the fixed commitment and the yardstick for performance. The forecast is a separate thing: closed actual months plus a best estimate for the rest of the year. Its example splits the gap between EBITDA budget and forecast by driver, such as volume, price and mix, energy and hiring timing, so each driver has an owner. Management actions stay outside the forecast until they have an owner, a timing and a delivery probability.
Each department's plan can pass and the total still fail
Corporate Finance AI describes a manufacturer that planned 15% sales growth, about 10% of it from volume. A months-long equipment shutdown pushed the inventory forecast negative, buying the gap externally cost more per tonne, HR had left out the operators the new line needed, and capex ran about 12% over. The fix was to change the assumptions across Sales, Production, Procurement, HR and Finance together. A flat percentage cut to every budget would not have addressed the cause.
A passing Checks sheet proved the arithmetic only
Corporate Finance AI recounts a team that built a five-year three-statement model in Claude in Excel for an investor financing. It had formulas only in the forecast, one scenario selector and a Checks sheet of fifteen live validations. A sceptical second review found Base costs above Worst on some lines, cash building up while the revolver was still drawn, and a capex swing nobody could explain. The lesson in the write-up is that green checks show the sums work, not that the assumptions are sound.
Apple moves its deal-making under the CFO
Proper Honest Tech reports that Steve Smith takes over Apple's M&A function from Adrian Perica, who moves into Eddy Cue's services organisation. M&A now reports to CFO Kevan Parekh instead of the CEO, the first time since 2019. The writer notes Apple rarely buys anything large, with Beats at $3 billion in 2014 still its biggest deal. The change could mean tighter financial discipline on deals, or preparation for bigger ones.
The Knowledge Brief
Billions and trillions get tossed around in finance headlines as if they sit next to each other. One explainer this week shows how far apart they really are.
A trillion is further from a billion than it sounds
A StarTalk explainer measures the two in seconds: a billion seconds is about 31 years, while a trillion seconds is about 31,000 years. It adds that a trillion dollars laid end to end would circle the Earth about 3,000 times, and that earning $10,000 an hour over 40 hours a week would take about 48,000 years to reach a trillion. Next time a budget or a valuation is quoted in trillions, these comparisons are a quick way to feel the scale.
Deep Space Digest
This week's space reading is about finding things that are hard to see. Some are planets, some are old light, and one may have been sitting in NASA data since 2018.
Euclid's Milky Way bulge image and the planet hunt
A YouTube video from Astrum covers a single 26-hour Euclid observation made in nine pointings on 23 March 2025. It produced the largest high-resolution visible-light image of the Milky Way's bulge, with more than 60 million stars across 4.8 square degrees. The video says this matters for planet hunting: microlensing finds planets regardless of size, heat or host star, and this image is the before snapshot the Roman telescope (launching early 2027, per the video) will need to confirm events.
The time the whole universe glowed
A YouTube video makes the case that the cosmic microwave background formed at about 3,000 (the unit goes unspoken, likely kelvin) and now sits near 3 K. Because the stretch in wavelength maps one to one to expansion, the video concludes the universe was about 1/1000 of its current size then, and that the early light was visible. The guests liken it to the surface of the sun, only found everywhere at once.
Where the search for alien life stands
A YouTube video runs through the main ways planets are found: transit, Doppler, direct imaging, astrometry and microlensing. It says over 6,000 exoplanets are known and there is still no Earth twin, and that finding one needs a coronagraph such as NASA's Habitable Worlds Observatory, or a Starshade, whose funding is on hold. It also proposes sampling Venus's cloud layer at about 50 km, where a PNA is stable to about 50 C in concentrated sulfuric acid.
A possible new planet in 2018 NASA data
The Rundown AI reports that independent researcher Pavel Rabtsevich used Claude Code and Codex to identify a potential new planet in NASA telescope data dating back to 2018. The report calls it only a potential planet, so it is unconfirmed. It is a reminder that old datasets can still hold surprises.
Tech in Translation
AI is now turning up serious security bugs and taking over more routine work. This week covers both, plus a few Google and iPhone changes that are useful to know.
AI model helps researchers find two critical PHP flaws
The Rundown AI reports that OpenAI's GPT-5.6 Sol helped researchers find two critical flaws in PHP, which sits behind about 70% of websites. Security firm Calif found the first bug on 1 August with the model, and Wiz caught both weeks later using Sol and Gemini 4 Argon. OpenAI says it worked with the PHP maintainers for two months and that several serious vulnerabilities are still being fixed.
Anthropic launches Claude Duet, a personal agent app
The Rundown AI says Anthropic has introduced Claude Duet, a standalone personal agent app powered by Opus 5.5. It can act proactively and keep working when your devices are closed, using its own cloud computer. Deletions, payments, signatures and outgoing messages need your sign-off first. It is a research preview for waitlisted US Max and Pro subscribers, rolling out in waves.
Let AI handle the copy and paste finishing tasks
The Rundown AI relays Nate Grahek's argument that any task that is only a copy and paste should go to an AI through a connector. His example is building a survey in Tally or Jotform and pulling the responses back into the chat. He suggests building trust in steps: read first, then write, then send, with AI drafting and a person pressing send.
An API key lets AI edit your site, with guardrails
Jeff Su says most paid tools offer an API alongside their screens, and a key lets an AI read your account, make a change and confirm it. He used one on a Ghost site to rewrite a page and update publish dates across 23 pages. His advice is to have the AI show the plan first, read the result back afterwards, give the smallest access the job needs, and revoke the key after large batches.
Google Docs on mobile can draft a document from your voice
Google Workspace Updates says Docs Live is rolling out to eligible Workspace users in the Google Docs apps on Android and iOS. It turns spoken conversation into a structured first draft and pulls context from Gmail, Chat and Drive. It is on by default where Gemini in Docs is enabled.
Gemini Notebook adds more control over slides and videos
According to Google Workspace Updates, Slide Decks and Video Overviews in Gemini Notebook now begin with a guided flow that suggests focus topics, formats and visual styles. A Review toggle lets you approve the outline or producer notes before anything is generated, and a quick first pass is still available. The rollout began on 16 October and takes up to 15 days.
Google Meet tile pairing comes to Android and iOS
Google Workspace Updates reports that Meet's tile pairing now works on mobile. It keeps a participant's video beside an interpreter or co-presenter and attributes live captions to both speakers, which helps Deaf and hard-of-hearing participants. Android got it from 12 October and iOS follows from 2 November. It is on by default, and admins can switch it off per OU.
iPhone tip: decide on Connectivity Assist
Proper Honest Tech's tip 324 covers Connectivity Assist, which lets an iPhone use mobile data alongside Wi-Fi when the connection struggles. On iOS 27 you find it under Settings, then Wi-Fi, and can switch it off if your data allowance is limited. The same screen shows how much mobile data it has used, with a Reset Data Usage button to restart the count.
