TMT Selected News

COMPANY RESULTS

TOWR (Buy, TP: Rp1,400) Inline Earnings Supported by Items Below EBITDA

4Q24 Earnings:

• TOWR reported net profit of Rp888bn (+5.5%qoq, +7.3%yoy), supported by stable revenue and positive contributions from items below the EBITDA line.
• Revenue was largely stable at Rp3.29tr (-0.3%qoq, +8.8%yoy), as robust non-tower revenue offset weaker tower revenue. Non-tower revenue contributed 33.8% of total revenue, increasing by 90 bps qoq.
• EBITDA was Rp2.75tr (-2.1%qoq, +7.5%yoy), translating to an EBITDA margin of 83.8% (-150 bps qoq). This lower annualized EBITDA resulted in increased ND/EBITDA leverage.
• EBT rose to Rp1.10tr (+7.9%qoq, +9.4%yoy), driven by favorable forex gains and increased other operating income.

 

FY24 Earnings:

• TOWR reported net profit of Rp3.34tr (+2.5% yoy), which was well inline with our estimate at 102.6% and broadly inline with consensus at 98.8%.
• Revenue grew to Rp12.74tr (+8.5%yoy), supported by strong non-tower revenue growth (+26.2%yoy), compensating for softer growth in tower revenue. Results were well inline with our and consensus estimates.
• EBITDA was resilient at Rp10.70tr (+7.2% yoy), meeting expectations.
• EBT of Rp4.14tr was inline with our estimates and broadly inline with consensus, amid higher D&A and financial expenses recorded in 2024.

 

The earnings call is scheduled for March 26th at 2pm. (Niko Margaronis & Kafi Ananta – BRIDS)

LOCAL NEWS

BUKA Plans Rp1.9Tr Share Buyback

PT Bukalapak.com Tbk (BUKA) has announced a share buyback plan worth up to Rp1.9tr to maintain stock price stability amid market volatility. The buyback will be carried out gradually or in one go through the Indonesia Stock Exchange or off-market transactions, within the period from March 26 to June 25, 2025. The company emphasized that this move aims to boost shareholder confidence and support long-term growth. The buyback will proceed without requiring approval from the General Meeting of Shareholders (GMS), in accordance with applicable OJK regulations. (IDX)

 

XLSmart Formed Following Shareholder Approval of XL Axiata and Smartfren Merger

25 March 2025,the EGMS (RUPSLB) has officially approved the merger between XL Axiata Smartfren . The merged entity will operate under the new name XLSmart. Alongside the approval of the merger, shareholders also finalized the composition of the new Board of Commissioners for XLSmart.

• President Commissioner: Arsjad Rasjid
• Commissioners: Vivek Sood, L. Krisnan Cahya, Nik Rizal Kamil, Sean Quek, David R. Dean
• Independent Commissioners: Retno Lestari Priansari Marsudi, Robert Pakpahan, Willem Lucas Timmerman (IDX)

 

EXCL to Distribute Rp1.12Tr in Dividends for FY24, Equivalent to Rp85.7 per Share

EXCL to distribute a cash dividend of up to Rp1.2tn for the 2024 fiscal year, representing a maximum of 62% of the company’s net profit, as per AGMS 25 March 2025. The cash dividend translates to Rp85.7 per share, and will be paid based on the following schedule:

• Cum Dividend in Regular and Negotiable Market: 11 April 2025
• Ex Dividend in Regular and Negotiable Market: 14 April 2025
• Cum Dividend in Cash Market: 15 April 2025
• Ex Dividend in Cash Market: 16 April 2025
• Recording Date: 15 April 2025
• Dividend Payment Date: 24 April 2025

The dividend implies a yield of ~3.8% based on the current share price. (IDX)

 

XLSmart Repositions Spectrum Strategy for 4G/5G After 900 MHz Cutback

XLSmart is committed to focus on expanding its 4G and 5G networks following Komdigi's decision to reclaim 2x7.5 MHz of 900 MHz spectrum previously held by XL Axiata. The ministry deemed the spectrum inefficient for future network development, particularly since 5G requires at least 100 MHz of contiguous bandwidth. Forcing 4G deployment on the limited 900 MHz band would leave XLSmart with fragmented spectrum, creating inefficiencies, especially as legacy 2G networks—where such leftover bandwidth might be used—are being phased out. (Bisnis)

 

AXIS Bronet Data Packages See Price Increases in March, Yield/GB Slightly Up

According to our latest monthly price tracker, AXIS (a brand under XL Axiata) has implemented several price adjustments across its Bronet data packages for March 2025. The majority of changes occurred in the 7-day and 1-day validity categories, with select packages in the 30-day tier also seeing a modest uptick. The average implied yield per GB increased from Rp3,458 in February to Rp3,807 in March. The uptick in pricing may reflect efforts to improve ARPU. (Company, BRIDS)

GoPay Launches Government Program Info Page on Its App

GoPay has launched a new "Government Programs" page within its app, providing users with easy access to information on various public initiatives aimed at supporting the wider community. The page includes programs such as Free Health Check-ups, Kartu Prakerja (Pre-Employment Card), Free Nutritious Meals, Kartu Indonesia Pintar (Smart Indonesia Card), QRIS payment services, and a public awareness campaign titled “Gambling Always Loses”, supported by Komdigi. (Bisnis)

 

Meta Launches AI Studio in Indonesia, Letting Users Create Custom AI Characters for Instagram and WhatsApp

Meta has officially launched AI Studio in Indonesia, a new feature that allows users to create and interact with custom AI characters based on their personal interests. Through AI Studio, accessible via ai.meta.com/ai-studio or the “AI Chat” feature on Instagram, users can design AI personas by customizing their name, personality, tone, avatar, and slogan. These AI characters can serve various roles, such as mentors, companions, or entertainers—and can be shared and engaged with across Meta platforms like Instagram and WhatsApp. (Katadata)

 

OTHER FOREIGN TRENDS

China Venture capitalists hold back on AI deals despite DeepSeek buzz; Asia's US$13.5bn AI invest in 2024 still lags US's US$101bn

DeepSeek may have shaken Wall Street and Silicon Valley into rethinking AI strategies, but the startup’s rise hasn’t spurred venture capital activity in China, where government funds remain the main backers. As of March 11, China recorded 100 AI and machine learning deals worth $917.8mn—well behind the $2.6bn from 196 US deals in Q1 2024.

Despite a broader VC slowdown since 2021, China led Asia in AI/ML investment last year with 715 deals totaling $7.3bn, far ahead of South Korea (308 deals, $1.8bn) and India (306 deals, $1.7bn). Asia’s AI lag behind the US stems partly from restrictions on foreign capital—especially US funds now barred from investing in China—prompting global firms to separate US-China operations. State involvement in Chinese AI firms like DeepSeek also adds regulatory hurdles, making a deal surge unlikely. India, however, saw strong AI/ML growth, rising to second in Asia. Still, Asia’s total AI investment of $13.5bn in 2024 lags far behind the US’s $100.5bn.(asia.Nikkei.com)

Texas private school’s AI tutor boosts test scores into top 2% nationally

Heritage Classical Academy, a private K-12 school in Houston, has integrated AI tutors (Merlyn Mind) into every classroom, leading to a dramatic rise in student test scores. The AI acts as a real-time teaching assistant, helping teachers tailor lessons, answer questions, and provide personalized instruction. Students benefit from instant feedback and enhanced independent learning. Key outcomes include faster grading, improved engagement, and top-tier standardized test performance. (Fox News)

 

Global sellside securities firm cuts AI server shipment forecasts (Mar 2025)

2025/2026 rack-level AI server estimates cut from 31k/66k units to 19k/57k, due to GPU transitions, demand uncertainty, production complexity, efficient models (e.g., DeepSeek), and tariff risks.

Financial Impact:

• Rack server market: Now seen at $54bn (2025) and $156bn (2026) (vs prior $88bn/$182bn).
• AI training server revenue: Cut to $160bn/$260bn (from $179bn/$248bn).
• AI inference servers: Strong growth expected.
• General servers: Modest gains from CPU upgrades.
• Outlook: Growth continues but is slowing. ASIC servers (servers using custom chips built for specific tasks like AI inference) may outperform GPU servers. Goldman stays positive on AI inference and general server demand. (Newsportal, Goldman Sachs)

 

Xiaomi Raises US$5.5Bn in Upsized Hong Kong Share Sale

Xiaomi Corp. has raised approximately US$5.5bn through an upsized share sale in Hong Kong, placing 800mn shares at HK$53.25 each, a 6.6% discount to the previous close, as the company accelerates investments in its fast-growing electric vehicle (EV) business. The deal, which drew strong interest from long-only investors and sovereign wealth funds, marks one of the largest equity raises in Hong Kong since 2021 and reflects renewed global investor confidence amid a 19% YTD gain in the Hang Seng Index. Proceeds will be used to expand EV production capacity, bolster R&D—including in AI—and support long-term deleveraging, although the short-term impact may be dilutive to existing shareholders. (YahooFinance)

 

Alibaba’s Tsai Warns of ‘Bubble’ in AI Data Center Buildout

Alibaba Chairman Joe Tsai has cautioned that a bubble may be forming in the global AI data center buildout, as tech giants and funds ramp up infrastructure spending without clear demand. Speaking at the HSBC Global Investment Summit in Hong Kong, Tsai criticized the “indiscriminate” rush to construct data centers, often without confirmed customers—warning that speculative investments could backfire. This comes as Alibaba plans to invest US$52bn in AI over the next three years, while U.S. tech rivals like Amazon, Alphabet, Meta, and Microsoft collectively commit over US$300bn to AI infrastructure. Despite Alibaba’s recent AI resurgence with its Qwen platform, Tsai remains skeptical of the sustainability of such aggressive global spending. (YahooFinance)

 

Alibaba says to restart hiring, sees signs of start of AI bubble in the US

Alibaba Chairman Joe Tsai announced that the company will resume hiring after 12 quarters of headcount decline, citing renewed business confidence following President Xi Jinping’s rare meeting with Chinese tech leaders, including Jack Ma. Speaking at HSBC’s Global Investment Summit in Hong Kong, Tsai praised the meeting as a clear signal for companies to reinvest and hire, viewing it as essential for restoring consumer confidence amid China’s sluggish economic recovery. At the same time, Tsai voiced concern over the massive AI infrastructure spending in the U.S., calling the scale of investment, reportedly up to US$500bn, a potential early sign of a bubble, especially with data centers being built without confirmed demand. While Alibaba plans to invest US$52bn in AI and cloud over the next three years, Tsai believes the real momentum lies in practical applications and measured growth, boosted by the rise of low-cost, open-source AI models like DeepSeek. (Reuters)

 

Growing trust in AI among finance leaders yet 80% lack confidence in planning beyond 6 months

72% of finance leaders would consider input from AI in making a business decision or recommendation but most companies are 2-3 years away from being able to fully maximize its value . Overall, security concerns, difficulty integrating AI into existing systems and data quality issues are ranked as the top three challenges to unlock AI’s full potential . While pressure grows on finance leaders to be more strategic, only 20% of businesses feel confident making plans more than 6 months out, and only 10% feel confident making plans for the next 2 years and beyond. (Pigment)

 

Sam Altman’s World Network in Talks With Visa for Stablecoin Payments Wallet

World Network, the blockchain ecosystem tied to Worldcoin and overseen by Tools for Humanity (cofounded by OpenAI CEO Sam Altman), is in talks with Visa to integrate Visa card functionality into its self-custody crypto wallet, World Wallet.

The goal is to make World Wallet a multifunctional mini bank account, enabling:

• On-chain Visa card features
• Stablecoin payments to Visa merchants
• FX and crypto trading
• Fiat deposit and withdrawal options

Tools for Humanity has already engaged with crypto card facilitators like Rain, which provides on-chain Visa cards for other blockchain projects (e.g., Optimism, Avalanche).

Crypto purists decry the move as a betrayal of decentralization, calling it a “centralized shitcoin with a Visa sticker.” (Reuters, SosMed)

 

China’s DeepSeek just released one of the most important open-source AI models to date.

The model, DeepSeek-V3-0324, quietly went live on March 24, 2025. It’s already drawing attention for combining top-tier performance with unprecedented accessibility.

Highlights:

• Fully open under MIT license — free for commercial and personal use
• 685B total parameters, but only 37B active at once via Mixture-of-Experts (MoE), making it highly efficient
• Includes Multi-Head Latent Attention (better context) and Multi-Token Prediction (faster output), boosting speed by up to 80%
• Runs locally on high-end consumer devices like Apple’s Mac Studio with M3 Ultra — no need for massive GPU clusters
• Beats Claude Sonnet 3.5 and other major models on key benchmarks like MMLU-Pro, LiveCodeBench, and MATH-500
• Now ranks as the best-performing non-reasoning model on the Artificial Analysis Intelligence Index

 

What’s even more remarkable is how this model challenges the dominant approach of Western AI labs. While companies like OpenAI and Anthropic release models behind paywalls, DeepSeek is pushing forward a fully open, community-accessible path — one that may accelerate global innovation.

Next up: DeepSeek-R2, a reasoning model expected to launch soon and potentially rival GPT-5.

Open access to its intelligence. Consumer-ready performance. Quiet disruption. DeepSeek is reshaping not just what AI can do — but anyone who gets to build with it. (Venturebeat)

To see the full version of this News, please click here